Merge with 4.0

BitKeeper/etc/gone: auto-union BitKeeper/etc/ignore: auto-union BitKeeper/etc/logging_ok: auto-union BitKeeper/deleted/.del-.my_sys.h.swp~f6a4a7f8dae03f18: Auto merged BitKeeper/etc/config: Auto merged acconfig.h: Auto merged acinclude.m4: Auto merged Docs/manual.texi: Auto merged client/mysqlcheck.c: Auto merged client/mysqlshow.c: Auto merged include/myisam.h: Auto merged include/violite.h: Auto merged isam/pack_isam.c: Auto merged libmysql/libmysql.c: Auto merged libmysqld/lib_sql.cc: Auto merged myisam/Makefile.am: Auto merged myisam/ft_nlq_search.c: Auto merged myisam/mi_open.c: Auto merged myisam/mi_write.c: Auto merged myisam/sort.c: Auto merged mysql-test/mysql-test-run.sh: Auto merged mysql-test/r/fulltext.result: Auto merged mysql-test/r/func_test.result: Auto merged mysql-test/r/isam.result: Auto merged mysql-test/r/rpl_replicate_do.result: Auto merged mysql-test/r/variables.result: Auto merged mysql-test/t/func_test.test: Auto merged mysql-test/t/myisam.test: Auto merged mysql-test/t/rpl_rotate_logs.test: Auto merged mysql-test/t/variables.test: Auto merged mysys/hash.c: Auto merged mysys/tree.c: Auto merged scripts/Makefile.am: Auto merged sql/Makefile.am: Auto merged sql/filesort.cc: Auto merged sql/gen_lex_hash.cc: Auto merged sql/ha_berkeley.cc: Auto merged sql/ha_innodb.cc: Auto merged sql/handler.cc: Auto merged sql/handler.h: Auto merged sql/hash_filo.h: Auto merged sql/hostname.cc: Auto merged sql/item_func.cc: Auto merged sql/item_func.h: Auto merged sql/item_timefunc.cc: Auto merged sql/lex.h: Auto merged sql/lock.cc: Auto merged sql/log.cc: Auto merged sql/log_event.h: Auto merged sql/mini_client.cc: Auto merged sql/opt_range.cc: Auto merged sql/opt_sum.cc: Auto merged sql/repl_failsafe.cc: Auto merged sql/set_var.h: Auto merged sql/slave.h: Auto merged sql/sql_acl.cc: Auto merged sql/sql_analyse.cc: Auto merged sql/sql_cache.cc: Auto merged sql/sql_insert.cc: Auto merged sql/sql_load.cc: Auto merged sql/sql_rename.cc: Auto merged sql/sql_repl.cc: Auto merged sql/sql_show.cc: Auto merged sql/sql_table.cc: Auto merged sql/sql_udf.cc: Auto merged sql/share/czech/errmsg.txt: Auto merged sql/share/danish/errmsg.txt: Auto merged sql/share/dutch/errmsg.txt: Auto merged sql/share/english/errmsg.txt: Auto merged sql/share/estonian/errmsg.txt: Auto merged sql/share/french/errmsg.txt: Auto merged sql/share/german/errmsg.txt: Auto merged sql/share/greek/errmsg.txt: Auto merged sql/share/hungarian/errmsg.txt: Auto merged sql/share/italian/errmsg.txt: Auto merged sql/share/japanese/errmsg.txt: Auto merged sql/share/korean/errmsg.txt: Auto merged sql/share/norwegian/errmsg.txt: Auto merged sql/table.h: Auto merged sql/unireg.cc: Auto merged sql-bench/server-cfg.sh: Auto merged sql/share/norwegian-ny/errmsg.txt: Auto merged sql/share/polish/errmsg.txt: Auto merged sql/share/portuguese/errmsg.txt: Auto merged sql/share/romanian/errmsg.txt: Auto merged sql/share/russian/errmsg.txt: Auto merged sql/share/slovak/errmsg.txt: Auto merged sql/share/spanish/errmsg.txt: Auto merged sql/share/swedish/errmsg.txt: Auto merged sql/share/ukrainian/errmsg.txt: Auto merged sql/ha_myisam.cc: Merge with 4.0 Removed some commented code sql/sql_db.cc: Merge with 4.0 Optimized the logging of the drop db call a bit sql/sql_update.cc: Added comment
2025-07-27 18:02:13 +03:00 · 2002-11-21 15:56:48 +02:00
parent 2723dbdb5e 70a17cd5a7
commit 33fc0d53b5
300 changed files with 7665 additions and 2426 deletions
--- a/.bzrignore
+++ b/.bzrignore
@ -31,8 +31,11 @@ BitKeeper/etc/gone
 BitKeeper/etc/level
 BitKeeper/etc/pushed
 BitKeeper/tmp/*
+BitKeeper/tmp/bkr3sAHD
+BitKeeper/tmp/gone
 COPYING
 COPYING.LIB
+Docs/#manual.texi#
 Docs/INSTALL-BINARY
 Docs/include.texi
 Docs/internals.info
@ -73,6 +76,9 @@ PENDING/*
 TAGS
 aclocal.m4
 autom4te.cache/*
+autom4te.cache/output.0
+autom4te.cache/requests
+autom4te.cache/traces.0
 bdb/README
 bdb/btree/btree_auto.c
 bdb/build_unix/*
@ -81,17 +87,21 @@ bdb/build_vxworks/db_int.h
 bdb/build_win32/db.h
 bdb/build_win32/db_archive.dsp
 bdb/build_win32/db_checkpoint.dsp
+bdb/build_win32/db_config.h
+bdb/build_win32/db_cxx.h
 bdb/build_win32/db_deadlock.dsp
 bdb/build_win32/db_dll.dsp
 bdb/build_win32/db_dump.dsp
 bdb/build_win32/db_int.h
 bdb/build_win32/db_java.dsp
 bdb/build_win32/db_load.dsp
+bdb/build_win32/db_perf.dsp
 bdb/build_win32/db_printlog.dsp
 bdb/build_win32/db_recover.dsp
 bdb/build_win32/db_stat.dsp
 bdb/build_win32/db_static.dsp
 bdb/build_win32/db_tcl.dsp
+bdb/build_win32/db_test.dsp
 bdb/build_win32/db_upgrade.dsp
 bdb/build_win32/db_verify.dsp
 bdb/build_win32/ex_access.dsp
@ -107,9 +117,16 @@ bdb/build_win32/excxx_lock.dsp
 bdb/build_win32/excxx_mpool.dsp
 bdb/build_win32/excxx_tpcb.dsp
 bdb/build_win32/include.tcl
+bdb/build_win32/libdb.def
 bdb/build_win32/libdb.rc
 bdb/db/crdel_auto.c
 bdb/db/db_auto.c
+bdb/dbinc_auto/*.*
+bdb/dbreg/dbreg_auto.c
+bdb/dist/autom4te.cache/*
+bdb/dist/autom4te.cache/output.0
+bdb/dist/autom4te.cache/requests
+bdb/dist/autom4te.cache/traces.0
 bdb/dist/config.hin
 bdb/dist/configure
 bdb/dist/tags
@ -118,11 +135,17 @@ bdb/dist/template/gen_client_ret
 bdb/dist/template/rec_btree
 bdb/dist/template/rec_crdel
 bdb/dist/template/rec_db
+bdb/dist/template/rec_dbreg
+bdb/dist/template/rec_fileops
 bdb/dist/template/rec_hash
 bdb/dist/template/rec_log
 bdb/dist/template/rec_qam
 bdb/dist/template/rec_txn
+bdb/examples_c/ex_apprec/ex_apprec_auto.c
+bdb/examples_c/ex_apprec/ex_apprec_auto.h
+bdb/examples_c/ex_apprec/ex_apprec_template
 bdb/examples_java
+bdb/fileops/fileops_auto.c
 bdb/hash/hash_auto.c
 bdb/include/btree_auto.h
 bdb/include/btree_ext.h
@ -151,16 +174,33 @@ bdb/include/tcl_ext.h
 bdb/include/txn_auto.h
 bdb/include/txn_ext.h
 bdb/include/xa_ext.h
+bdb/java/src/com/sleepycat/db/Db.java
+bdb/java/src/com/sleepycat/db/DbBtreeStat.java
 bdb/java/src/com/sleepycat/db/DbConstants.java
+bdb/java/src/com/sleepycat/db/DbHashStat.java
+bdb/java/src/com/sleepycat/db/DbLockStat.java
+bdb/java/src/com/sleepycat/db/DbLogStat.java
+bdb/java/src/com/sleepycat/db/DbMpoolFStat.java
+bdb/java/src/com/sleepycat/db/DbQueueStat.java
+bdb/java/src/com/sleepycat/db/DbRepStat.java
+bdb/java/src/com/sleepycat/db/DbTxnStat.java
+bdb/libdb_java/java_stat_auto.c
+bdb/libdb_java/java_stat_auto.h
 bdb/log/log_auto.c
 bdb/qam/qam_auto.c
 bdb/rpc_client/db_server_clnt.c
 bdb/rpc_client/gen_client.c
+bdb/rpc_server/c/db_server_proc.c
+bdb/rpc_server/c/db_server_proc.sed
+bdb/rpc_server/c/db_server_svc.c
+bdb/rpc_server/c/db_server_xdr.c
+bdb/rpc_server/c/gen_db_server.c
 bdb/rpc_server/db_server.x
 bdb/rpc_server/db_server_proc.sed
 bdb/rpc_server/db_server_svc.c
 bdb/rpc_server/db_server_xdr.c
 bdb/rpc_server/gen_db_server.c
+bdb/test/TESTS
 bdb/test/include.tcl
 bdb/test/logtrack.list
 bdb/txn/txn_auto.c
@ -194,6 +234,7 @@ configure.lineno
 core
 db-*.*.*
 dbug/user.t
+depcomp
 extra/comp_err
 extra/my_print_defaults
 extra/mysql_install
@ -209,6 +250,9 @@ include/my_global.h
 include/mysql_version.h
 include/widec.h
 innobase/autom4te.cache/*
+innobase/autom4te.cache/output.0
+innobase/autom4te.cache/requests
+innobase/autom4te.cache/traces.0
 innobase/configure.lineno
 innobase/conftest.s1
 innobase/conftest.subs
@ -392,9 +436,11 @@ mysql.kdevprj
 mysql.proj
 mysqld.S
 mysqld.sym
+mysys/#mf_iocache.c#
 mysys/charset2html
 mysys/getopt.c
 mysys/getopt1.c
+mysys/ste5KbMa
 mysys/test_charset
 mysys/test_dir
 mysys/test_io_cache
@ -511,33 +557,3 @@ vio/test-ssl
 vio/test-sslclient
 vio/test-sslserver
 vio/viotest-ssl
-bdb/build_win32/db_config.h
-bdb/build_win32/db_cxx.h
-bdb/build_win32/db_perf.dsp
-bdb/build_win32/db_test.dsp
-bdb/build_win32/libdb.def
-bdb/dbinc_auto/*.*
-bdb/dbreg/dbreg_auto.c
-bdb/dist/template/rec_dbreg
-bdb/dist/template/rec_fileops
-bdb/examples_c/ex_apprec/ex_apprec_auto.c
-bdb/examples_c/ex_apprec/ex_apprec_auto.h
-bdb/examples_c/ex_apprec/ex_apprec_template
-bdb/fileops/fileops_auto.c
-bdb/java/src/com/sleepycat/db/Db.java
-bdb/java/src/com/sleepycat/db/DbBtreeStat.java
-bdb/java/src/com/sleepycat/db/DbHashStat.java
-bdb/java/src/com/sleepycat/db/DbLockStat.java
-bdb/java/src/com/sleepycat/db/DbLogStat.java
-bdb/java/src/com/sleepycat/db/DbMpoolFStat.java
-bdb/java/src/com/sleepycat/db/DbQueueStat.java
-bdb/java/src/com/sleepycat/db/DbRepStat.java
-bdb/java/src/com/sleepycat/db/DbTxnStat.java
-bdb/libdb_java/java_stat_auto.c
-bdb/libdb_java/java_stat_auto.h
-bdb/rpc_server/c/db_server_proc.c
-bdb/rpc_server/c/db_server_proc.sed
-bdb/rpc_server/c/db_server_svc.c
-bdb/rpc_server/c/db_server_xdr.c
-bdb/rpc_server/c/gen_db_server.c
-bdb/test/TESTS
--- a/BUILD/Makefile.am
+++ b/BUILD/Makefile.am
@ -37,7 +37,7 @@ EXTRA_DIST =		FINISH.sh \
 			compile-pentium-pgcc \
 			compile-solaris-sparc \
 			compile-solaris-sparc-debug \
-			compile-solaris-sparc-fortre \
+			compile-solaris-sparc-forte \
 			compile-solaris-sparc-purify

 # Don't update the files from bitkeeper
--- a/BUILD/compile-pentium-valgrind-max
+++ b/BUILD/compile-pentium-valgrind-max
@ -0,0 +1,13 @@
+#! /bin/sh
+
+path=`dirname $0`
+. "$path/SETUP.sh"
+
+extra_flags="$pentium_cflags $debug_cflags -DHAVE_purify"
+c_warnings="$c_warnings $debug_extra_warnings"
+cxx_warnings="$cxx_warnings $debug_extra_warnings"
+extra_configs="$pentium_configs $debug_configs"
+
+extra_configs="$extra_configs --with-berkeley-db --with-innodb --with-embedded-server --with-openssl"
+
+. "$path/FINISH.sh"
--- a/BUILD/compile-solaris-sparc-forte
+++ b/BUILD/compile-solaris-sparc-forte
@ -0,0 +1,39 @@
+#! /bin/sh
+
+gmake -k clean || true
+/bin/rm -f */.deps/*.P config.cache
+ 
+aclocal && autoheader && aclocal && automake && autoconf
+(cd bdb/dist && sh s_all)
+(cd innobase && aclocal && autoheader && aclocal && automake && autoconf)
+if [ -d gemini ]
+then
+   (cd gemini && aclocal && autoheader && aclocal && automake && autoconf)
+fi
+
+
+# Assume Forte is installed in /opt/SUNWSpro
+
+PATH=/opt/SUNWspro/bin/:$PATH
+
+# For "optimal" code for this computer add -fast to EXTRA
+# To compile 64 bit, add -xarch=v9 to EXTRA_64_BIT
+
+EXTRA_64_BIT="-xarch=v9"	# Remove comment to get 64 bit binaries
+EXTRA="-fast"			# Remove comment to target current machine
+
+#
+# The following should not need to be touched
+#
+
+STD="-mt -D_FORTEC_ $EXTRA $EXTRA_64_BIT"
+ASFLAGS="$EXTRA_64_BIT" \
+CC=cc-5.0 CFLAGS="-Xa -xstrconst $STD" \
+CXX=CC CXXFLAGS="-noex $STD" \
+./configure --prefix=/usr/local/mysql --enable-assembler --with-extra-charsets=complex --enable-thread-safe-client
+
+gmake -j 4
+if [ $? = 0 ]
+then
+  make test
+fi
--- a/BUILD/compile-solaris-sparc-fortre
+++ b/BUILD/compile-solaris-sparc-fortre
@ -1,19 +0,0 @@
-#! /bin/sh
-
-gmake -k clean || true
-/bin/rm -f */.deps/*.P config.cache
- 
-aclocal && autoheader && aclocal && automake && autoconf
-(cd bdb/dist && sh s_all)
-(cd innobase && aclocal && autoheader && aclocal && automake && autoconf)
-if [ -d gemini ]
-then
-   (cd gemini && aclocal && autoheader && aclocal && automake && autoconf)
-fi
-
-PATH=/opt/SUNWspro/bin/:$PATH
-CC=cc CFLAGS="-Xa -fast -xO4 -native -xstrconst -mt -D_FORTREC_" \
-CXX=CC CXXFLAGS="-noex -xO4 -mt"  \
-./configure --prefix=/usr/local/mysql --enable-assembler --with-extra-charsets=complex --enable-thread-safe-client
-
-gmake -j 4
--- a/BitKeeper/etc/config
+++ b/BitKeeper/etc/config
@ -42,7 +42,7 @@ single_host:
 # discovers a problem which requires local intervention.  Please make the
 # contact information accurate so we can support you.
 # 
-contact: Sasha Pachev <sasha@mysql.com> 
+contact: sys@mysql.com
 # 
 # It is very important that this email address is filled out and accurate.
 # If someone converts your repository to open logging (which you may not
@ -51,7 +51,7 @@ contact: Sasha Pachev <sasha@mysql.com>
 # response from anyone else at your location after 90 days, then open logging
 # will be implicitly approved.
 # 
-email: sasha@mysql.com 
+email: sys@mysql.com 
 # 
 # Add your street address if you like, it is optional.
 # 
--- a/BitKeeper/etc/gone
+++ b/BitKeeper/etc/gone
@ -199,7 +199,9 @@ BK|sql-bench/Results/ATIS-mysql-3.21-Linux_2.2.1_i686|19700101030959|02022|660fb
 BK|sql-bench/Results/ATIS-mysql-Linux_2.2.10_i686|19700101030959|02025|3fa4d167cceff7e8
 BK|sql-bench/Results/ATIS-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02312|84ca3b85ff306133
 BK|sql-bench/Results/ATIS-mysql-Linux_2.2.14_i686_xeon|19700101030959|02044|3e820c28bf4af63a
+BK|sql-bench/Results/ATIS-mysql-SunOS_5.5.1_sun4u|19700101030959|02031|dfb4c5f6b6db3b49
 BK|sql-bench/Results/ATIS-mysql-SunOS_5.6_sun4m|19700101030959|02032|62028e0375b3b8b
+BK|sql-bench/Results/ATIS-mysql-SunOS_5.7_sun4u|19700101030959|02034|be0d9789776c5ed7
 BK|sql-bench/Results/ATIS-mysql_3.21-Linux_2.0.35_i686|19700101030959|02036|c25425e045ca8dfc
 BK|sql-bench/Results/ATIS-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02304|cbe120d860296d2f
 BK|sql-bench/Results/ATIS-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02027|a74e7b82d3908fa9
@ -263,7 +265,9 @@ BK|sql-bench/Results/RUN-mysql-3.21-Linux_2.2.1_i686|19700101030959|02050|f6fdd6
 BK|sql-bench/Results/RUN-mysql-Linux_2.2.10_i686|19700101030959|02041|712f52be5d195406
 BK|sql-bench/Results/RUN-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02038|8ee87b26b91c86fe
 BK|sql-bench/Results/RUN-mysql-Linux_2.2.14_i686_xeon|19700101030959|02055|17854e751e1d9d1d
+BK|sql-bench/Results/RUN-mysql-SunOS_5.5.1_sun4u|19700101030959|02058|afbba182428e20df
 BK|sql-bench/Results/RUN-mysql-SunOS_5.6_sun4m|19700101030959|02059|eafc8188345e262b
+BK|sql-bench/Results/RUN-mysql-SunOS_5.7_sun4u|19700101030959|02061|86e1dc0e25a8b8f
 BK|sql-bench/Results/RUN-mysql_3.21-Linux_2.0.35_i686|19700101030959|02064|ea8672d8473435
 BK|sql-bench/Results/RUN-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02310|a902e1a967d79c42
 BK|sql-bench/Results/RUN-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02030|413ab3b8a99e61e9
@ -273,7 +277,9 @@ BK|sql-bench/Results/alter-table-mysql-3.21-Linux_2.2.1_i686|19700101030959|0207
 BK|sql-bench/Results/alter-table-mysql-Linux_2.2.10_i686|19700101030959|02081|93b78a85b720a186
 BK|sql-bench/Results/alter-table-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02314|4ae4b989301df98b
 BK|sql-bench/Results/alter-table-mysql-Linux_2.2.14_i686_xeon|19700101030959|02057|64cc4b874cd6fabf
+BK|sql-bench/Results/alter-table-mysql-SunOS_5.5.1_sun4u|19700101030959|02087|9d7e75667fcb29ec
 BK|sql-bench/Results/alter-table-mysql-SunOS_5.6_sun4m|19700101030959|02088|8a1bd6589a189890
+BK|sql-bench/Results/alter-table-mysql-SunOS_5.7_sun4u|19700101030959|02090|ce74c2f623d3bb3
 BK|sql-bench/Results/alter-table-mysql_3.21-Linux_2.0.35_i686|19700101030959|02092|762639f2560976bd
 BK|sql-bench/Results/alter-table-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02316|1390155aad5b6e86
 BK|sql-bench/Results/alter-table-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02317|9090bebb62ef164b
@ -283,7 +289,9 @@ BK|sql-bench/Results/big-tables-mysql-3.21-Linux_2.2.1_i686|19700101030959|02106
 BK|sql-bench/Results/big-tables-mysql-Linux_2.2.10_i686|19700101030959|02109|99daa1c5370d077d
 BK|sql-bench/Results/big-tables-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02315|2804ec3c95be436a
 BK|sql-bench/Results/big-tables-mysql-Linux_2.2.14_i686_xeon|19700101030959|02074|290c2c3de9d8e6b
+BK|sql-bench/Results/big-tables-mysql-SunOS_5.5.1_sun4u|19700101030959|02115|7d7b6c0bf58b9b79
 BK|sql-bench/Results/big-tables-mysql-SunOS_5.6_sun4m|19700101030959|02116|f351a7f3e1e2257e
+BK|sql-bench/Results/big-tables-mysql-SunOS_5.7_sun4u|19700101030959|02118|ebc379b231312bbe
 BK|sql-bench/Results/big-tables-mysql_3.21-Linux_2.0.35_i686|19700101030959|02120|190e827e569c99a4
 BK|sql-bench/Results/big-tables-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02318|c5eabcb89ceac698
 BK|sql-bench/Results/big-tables-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02319|856d503725373684
@ -293,7 +301,9 @@ BK|sql-bench/Results/connect-mysql-3.21-Linux_2.2.1_i686|19700101030959|02134|c0
 BK|sql-bench/Results/connect-mysql-Linux_2.2.10_i686|19700101030959|02137|c92505d77e19d5ec
 BK|sql-bench/Results/connect-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02084|e7e2959b7387251f
 BK|sql-bench/Results/connect-mysql-Linux_2.2.14_i686_xeon|19700101030959|02071|ea19dc3ec55b3618
+BK|sql-bench/Results/connect-mysql-SunOS_5.5.1_sun4u|19700101030959|02142|a9493110fe62e0b1
 BK|sql-bench/Results/connect-mysql-SunOS_5.6_sun4m|19700101030959|02143|a10e3ddfa26a3e7f
+BK|sql-bench/Results/connect-mysql-SunOS_5.7_sun4u|19700101030959|02145|c67beb9e9d2cf32e
 BK|sql-bench/Results/connect-mysql_3.21-Linux_2.0.35_i686|19700101030959|02146|650abd213e6828c6
 BK|sql-bench/Results/connect-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02320|ce69cc65bc827b5c
 BK|sql-bench/Results/connect-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02066|f801e08429a4f7c6
@ -303,7 +313,9 @@ BK|sql-bench/Results/create-mysql-3.21-Linux_2.2.1_i686|19700101030959|02158|515
 BK|sql-bench/Results/create-mysql-Linux_2.2.10_i686|19700101030959|02161|9e7822f66df6aa76
 BK|sql-bench/Results/create-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02102|34ded91c5fc25de9
 BK|sql-bench/Results/create-mysql-Linux_2.2.14_i686_xeon|19700101030959|02139|50d15991293030ef
+BK|sql-bench/Results/create-mysql-SunOS_5.5.1_sun4u|19700101030959|02166|bbb5de66fc56de7b
 BK|sql-bench/Results/create-mysql-SunOS_5.6_sun4m|19700101030959|02221|9233114ae6f8c5f
+BK|sql-bench/Results/create-mysql-SunOS_5.7_sun4u|19700101030959|02223|7ee13bfcafeab498
 BK|sql-bench/Results/create-mysql_3.21-Linux_2.0.35_i686|19700101030959|02225|df1b037d17b33587
 BK|sql-bench/Results/create-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02321|e985e71d552ff09e
 BK|sql-bench/Results/create-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02099|483dcf223d5abf81
@ -313,7 +325,9 @@ BK|sql-bench/Results/insert-mysql-3.21-Linux_2.2.1_i686|19700101030959|02239|fd0
 BK|sql-bench/Results/insert-mysql-Linux_2.2.10_i686|19700101030959|02242|763edf9aec633f51
 BK|sql-bench/Results/insert-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02130|5be3d6f299738a31
 BK|sql-bench/Results/insert-mysql-Linux_2.2.14_i686_xeon|19700101030959|02141|c683ee4b9d214298
+BK|sql-bench/Results/insert-mysql-SunOS_5.5.1_sun4u|19700101030959|02247|8a9ae41f9a79f79
 BK|sql-bench/Results/insert-mysql-SunOS_5.6_sun4m|19700101030959|02248|3402d060ae20e19
+BK|sql-bench/Results/insert-mysql-SunOS_5.7_sun4u|19700101030959|02250|78efa132c6e252b9
 BK|sql-bench/Results/insert-mysql_3.21-Linux_2.0.35_i686|19700101030959|02252|60c0965dff31db07
 BK|sql-bench/Results/insert-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02322|ed252140ff399961
 BK|sql-bench/Results/insert-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02114|29a3b8a1ca8aa9d
@ -323,7 +337,9 @@ BK|sql-bench/Results/select-mysql-3.21-Linux_2.2.1_i686|19700101030959|02265|ed3
 BK|sql-bench/Results/select-mysql-Linux_2.2.10_i686|19700101030959|02268|a2e264d777b787d
 BK|sql-bench/Results/select-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02227|308117295c3bc096
 BK|sql-bench/Results/select-mysql-Linux_2.2.14_i686_xeon|19700101030959|02152|ead3f11b46ac626f
+BK|sql-bench/Results/select-mysql-SunOS_5.5.1_sun4u|19700101030959|02273|c9a1a498a052e268
 BK|sql-bench/Results/select-mysql-SunOS_5.6_sun4m|19700101030959|02274|4da215905bce988d
+BK|sql-bench/Results/select-mysql-SunOS_5.7_sun4u|19700101030959|02276|632c92971c61e34a
 BK|sql-bench/Results/select-mysql_3.21-Linux_2.0.35_i686|19700101030959|02278|5fadbac5f98696a
 BK|sql-bench/Results/select-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02323|e8c0871a668a610d
 BK|sql-bench/Results/select-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02127|963a98ed526e2be4
@ -333,7 +349,9 @@ BK|sql-bench/Results/wisconsin-mysql-3.21-Linux_2.2.1_i686|19700101030959|02290|
 BK|sql-bench/Results/wisconsin-mysql-Linux_2.2.10_i686|19700101030959|02288|301a82b12a84922b
 BK|sql-bench/Results/wisconsin-mysql-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02280|d01900af34fb33b8
 BK|sql-bench/Results/wisconsin-mysql-Linux_2.2.14_i686_xeon|19700101030959|02154|7525b23938631801
+BK|sql-bench/Results/wisconsin-mysql-SunOS_5.5.1_sun4u|19700101030959|02297|379705afa2e12378
 BK|sql-bench/Results/wisconsin-mysql-SunOS_5.6_sun4m|19700101030959|02298|ec61b14072715dc8
+BK|sql-bench/Results/wisconsin-mysql-SunOS_5.7_sun4u|19700101030959|02300|f27927f8c64ea8ad
 BK|sql-bench/Results/wisconsin-mysql_3.21-Linux_2.0.35_i686|19700101030959|02302|31703d40ea6b4f66
 BK|sql-bench/Results/wisconsin-mysql_fast-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02324|ec075a89dbdbbe6a
 BK|sql-bench/Results/wisconsin-pg-Linux_2.2.14_5.0_i686-cmp-mysql,pg|19700101030959|02325|233d5aa529979990
@ -727,6 +745,8 @@ mwagner@evoq.home.mwagner.org|mysql-test/xml/tests/sel000029.xml|20001017133713|
 mwagner@evoq.home.mwagner.org|mysql-test/xml/tests/sel000030.xml|20001017133600|63205|c2b25781eefaee9
 mwagner@evoq.home.mwagner.org|mysql-test/xml/xsl/README|20001013051514|26509|cd4bb681e5a0cd10
 mwagner@evoq.home.mwagner.org|mysql-test/xml/xsl/mysqltest.xsl|20001013051514|27425|1b8f6ec4f1b5f634
+mwagner@work.mysql.com|mysql-test/r/3.23/sel000001.result|20001010091454|28284|383913ae4505ec86
+mwagner@work.mysql.com|mysql-test/r/3.23/sel000002.result|20001010091454|29230|d1787e6fd5dbc1cc
 nick@nick.leippe.com|mysql-test/r/rpl_empty_master_crash.result|20020531235552|47718|615f521be2132141
 nick@nick.leippe.com|mysql-test/t/rpl_empty_master_crash.test|20020531235552|52328|99464e737639ccc6
 sasha@mysql.sashanet.com|BitKeeper/etc/logging_ok|20000801000905|12967|5b7d847a2158554
@ -734,19 +754,36 @@ sasha@mysql.sashanet.com|build-tags|20011125054855|05181|7afb7e785b80f97
 sasha@mysql.sashanet.com|build-tags|20011201050944|25384|b6f6fff142121618
 sasha@mysql.sashanet.com|libmysql_r/acconfig.h|20001128060846|51084|65f1202b3b5c345f
 sasha@mysql.sashanet.com|mysql-test/README.gcov|20001012045950|28177|5a6da067a30780ce
+sasha@mysql.sashanet.com|mysql-test/README.gcov|20001214012355|41825|2de7575ca81155e5
 sasha@mysql.sashanet.com|mysql-test/README|20001010001022|12739|108667adaeabe3f5
 sasha@mysql.sashanet.com|mysql-test/r/3.23/alt000001.result|20001122072330|24729|393103dbf15f35c9
+sasha@mysql.sashanet.com|mysql-test/r/3.23/ins000001.result|20001018175743|49824|f45c599efdf8352b
 sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000001.a.result|20001118063528|39426|2987b17db06808c3
 sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000001.b.result|20001118063528|44057|62e1fa91167cacc3
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000002.result|20001118063528|46039|109f5ceed1e0d64
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000003.result|20001118063528|48148|68d6ee00beaa011
 sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000004.a.result|20001118063528|50132|3415f066cb91c460
 sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000004.b.result|20001118063528|52094|352b35351551485
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000005.result|20001118063528|54071|a50962bc2340ab9a
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000006.result|20001118063528|56081|5653051e8ce6b4aa
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000007.result|20001121063807|21606|e0c3b6134e0884da
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000008.result|20001121063807|23636|c5cfee19ca5a7da9
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000009.result|20001121063807|25633|ed8042446ab97926
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000010.result|20001122072330|29430|3228109b8965b0f8
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000011.result|20001125024912|48851|c29dce30aa97f265
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000012.result|20001126062901|05938|35d6596da7b90fc5
 sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000012.status.result|20001126062901|09395|bbbd650b5beea32f
+sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000013.result|20001202171150|03876|ac5024e6cf6daac6
 sasha@mysql.sashanet.com|mysql-test/r/3.23/rpl000013.status.result|20001202171150|06069|6bee190c298cc9fd
+sasha@mysql.sashanet.com|mysql-test/r/3.23/sel000003.result|20001011230020|64653|d7b657b1e3a286a7
+sasha@mysql.sashanet.com|mysql-test/r/3.23/sel000100.res|20001205131218|23520|84ed46856cb3a69f
 sasha@mysql.sashanet.com|mysql-test/r/3.23/shw000001.result|20001121234128|16652|8b20b03d8319b9a5
 sasha@mysql.sashanet.com|mysql-test/r/binlog-backup-restore.result|20010424233926|16010|605de78abda64d27
 sasha@mysql.sashanet.com|mysql-test/r/df_crash.result|20010406010433|59989|4a3dbee64843953d
 sasha@mysql.sashanet.com|mysql-test/r/identity.result|20010910233028|16331|e41453a364242503
 sasha@mysql.sashanet.com|mysql-test/r/mrg000002.result|20001212152450|11492|745be0854aaaaf5e
+sasha@mysql.sashanet.com|mysql-test/r/slave-running.result|20001208141122|24303|f73e49462cf59e1f
+sasha@mysql.sashanet.com|mysql-test/r/slave-stopped.result|20001208141122|28916|25c134b1a4f1993a
 sasha@mysql.sashanet.com|mysql-test/std_data/m.MRG|20001212152450|17736|3f5632c37af00f18
 sasha@mysql.sashanet.com|mysql-test/std_data/m.frm|20001212152450|13897|e351dfe0b6824c0c
 sasha@mysql.sashanet.com|mysql-test/std_data/select-key.master|20001009234916|07315|e6b83af25df0ce5
--- a/BitKeeper/etc/logging_ok
+++ b/BitKeeper/etc/logging_ok
@ -16,6 +16,7 @@ bar@gw.udmsearch.izhnet.ru
 bell@sanja.is.com.ua
 bk@admin.bk
 davida@isil.mysql.com
+gluh@gluh.(none)
 heikki@donna.mysql.fi
 heikki@hundin.mysql.fi
 heikki@rescue.
@ -41,6 +42,7 @@ miguel@hegel.local
 miguel@light.
 miguel@light.local
 monty@bitch.mysql.fi
+monty@butch.
 monty@donna.mysql.fi
 monty@hundin.mysql.fi
 monty@mashka.mysql.fi
--- a/Build-tools/Do-compile
+++ b/Build-tools/Do-compile
@ -8,13 +8,13 @@ use Getopt::Long;
 $opt_distribution=$opt_user=$opt_config_env="";
 $opt_dbd_options=$opt_perl_options=$opt_config_options=$opt_make_options=$opt_suffix="";
 $opt_tmp=$opt_version_suffix="";
-$opt_help=$opt_delete=$opt_debug=$opt_stage=$opt_no_test=$opt_no_perl=$opt_with_low_memory=$opt_fast_benchmark=$opt_static_client=$opt_static_server=$opt_static_perl=$opt_sur=$opt_with_small_disk=$opt_local_perl=$opt_tcpip=$opt_build_thread=$opt_use_old_distribution=$opt_enable_shared=$opt_no_crash_me=$opt_no_strip=0;
+$opt_help=$opt_delete=$opt_debug=$opt_stage=$opt_no_test=$opt_no_perl=$opt_with_low_memory=$opt_fast_benchmark=$opt_static_client=$opt_static_server=$opt_static_perl=$opt_sur=$opt_with_small_disk=$opt_local_perl=$opt_tcpip=$opt_build_thread=$opt_use_old_distribution=$opt_enable_shared=$opt_no_crash_me=$opt_no_strip=$opt_with_debug=0;
 $opt_innodb=$opt_bdb=$opt_raid=$opt_libwrap=0;

 GetOptions(
 	"bdb",
 	"build-thread=i",
-	"config-env=s",
+	"config-env=s" => \@config_env,
 	"config-options=s" => \@config_options,
 	"dbd-options=s",
 	"debug",
@ -45,6 +45,7 @@ GetOptions(
 	"use-old-distribution",
 	"user=s",
 	"version-suffix=s",
+	"with-debug",
 	"with-low-memory",
 	"with-other-libc=s",
 	"with-small-disk",
@ -53,11 +54,6 @@ GetOptions(
 usage() if ($opt_help);
 usage() if (!$opt_distribution);

-if ($opt_bdb && $opt_version_suffix eq "")
-{
-  $opt_version_suffix="-max";
-}
-
 if (@make_options > 0)
 {
 	chomp(@make_options);
@ -70,6 +66,12 @@ if (@config_options > 0)
 	$opt_config_options= join(" ", @config_options);
 }

+if (@config_env > 0)
+{
+	chomp(@config_env);
+	$opt_config_env= join(" ", @config_env);
+}
+
 chomp($host=`hostname`);
 $full_host_name=$host;
 $connect_option= ($opt_tcpip ? "--host=$host" : "");
@ -208,6 +210,7 @@ if ($opt_stage <= 1)
  $opt_config_options.= " --disable-shared" if (!$opt_enable_shared); # Default for binary versions
  $opt_config_options.= " --with-berkeley-db" if ($opt_bdb);
  $opt_config_options.= " --with-client-ldflags=-all-static" if ($opt_static_client);
+	$opt_config_options.= " --with-debug" if ($opt_with_debug);
  $opt_config_options.= " --with-libwrap" if ($opt_libwrap);
  $opt_config_options.= " --with-low-memory" if ($opt_with_low_memory);
  $opt_config_options.= " --with-mysqld-ldflags=-all-static" if ($opt_static_server);
@ -258,7 +261,7 @@ if ($opt_stage <= 3)
  log_system("rm -fr mysql-3* mysql-4* $pwd/$host/*.tar.gz");
  log_system("nm -n sql/mysqld | gzip -9 -v 2>&1 > sql/mysqld.sym.gz | cat");

-  $flags.= "--no-strip" if ($opt_no_strip);
+  $flags.= "--no-strip" if ($opt_no_strip || $opt_with_debug);
  check_system("scripts/make_binary_distribution --tmp=$opt_tmp --suffix=$opt_suffix $flags",".tar.gz created");
  safe_system("mv mysql*.tar.gz $pwd/$host");
  if (-f "client/.libs/mysqladmin")
@ -381,13 +384,15 @@ if ($opt_stage <= 9 && !$opt_no_test)
  log_system("rm -f output/*");
  $tmp= $opt_fast_benchmark ? "--fast --user root --small-test" : "";
  check_system("perl ./run-all-tests --log --die-on-errors $connect_option $tmp","RUN-mysql");
+	# Run additional fast test with dynamic-row tables
+  check_system("perl ./run-all-tests --log --suffix=\"_dynamic_rows\" --die-on-errors $connect_option --fast --user=root --small-test --create-options=\"row_format=dynamic\"","RUN-mysql");
  if ($opt_innodb)
  {
-    check_system("perl ./run-all-tests --log --suffix=\"_innodb\" --die-on-errors $connect_option $tmp --create-option=\"type=innodb\"","RUN-mysql");
+    check_system("perl ./run-all-tests --log --suffix=\"_innodb\" --die-on-errors $connect_option $tmp --create-options=\"type=innodb\"","RUN-mysql");
  }
  if ($opt_bdb)
  {
-    check_system("perl ./run-all-tests --log --suffix=\"_bdb\" --die-on-errors $connect_option $tmp --create-option=\"type=bdb\"","RUN-mysql");
+    check_system("perl ./run-all-tests --log --suffix=\"_bdb\" --die-on-errors $connect_option $tmp --create-options=\"type=bdb\"","RUN-mysql");
  }
 }

@ -498,6 +503,9 @@ If user is empty then no mail is sent.
 --version-suffix suffix
 Set name suffix (e.g. 'com' or '-max') for a distribution

+--with-debug
+Build binaries with debug information (implies "--no-strip")
+
 --with-low-memory
 Use less memory when compiling.

@ -528,7 +536,7 @@ sub abort
    print TMP "To: $email\n";
    print TMP "Subject: $ver$opt_version_suffix compilation failed\n\n";
    close TMP;
-    system("tail -40 $log > $log.mail");
+    system("tail -n 40 $log > $log.mail");
    system("cat $mail_header_file $log.mail | $sendmail -t -f $email");
    unlink($mail_header_file);
    unlink("$log.mail");
@ -604,7 +612,7 @@ sub which
  my(@progs)=@_;
  foreach $prog (@progs)
  {
-    chomp($found=`which $prog | head -1`);
+    chomp($found=`which $prog | head -n 1`);
    if ($? == 0 && $found ne "" && index($found," ") == -1)
    {
      $found =~ s|/+|/|g;	# Make nicer output
--- a/Docs/internals.texi
+++ b/Docs/internals.texi
--- a/Docs/manual.texi
+++ b/Docs/manual.texi
@ -8202,6 +8202,10 @@ The following startup variables/options have been renamed:
 The startup options @code{record_buffer}, @code{sort_buffer} and
@code{warnings} will still work in MySQL 4.0 but are deprecated.
@item
+The mysqld option @code{--safe_show_database} doesn't work anymore.  One
+should instead give the @code{SHOW DATABASES} privileges to everyone that
+need to see all databases.
+@item
 The following SQL variables have changed name.
@c arjen note: New table, not yet measured for O'Reilly/DocBook.
@multitable @columnfractions .50 .50
@ -19879,7 +19883,6 @@ differ somewhat:
 | query_cache_limit               | 1048576                      |
 | query_cache_size                | 0                            |
 | query_cache_type                | ON                           |
-| safe_show_database              | OFF                          |
 | server_id                       | 0                            |
 | slave_net_timeout               | 3600                         |
 | skip_external_locking           | ON                           |
@ -20314,7 +20317,8 @@ This may be set (only numeric) to
 Don't show databases for which the user doesn't have any database or
 table privileges. This can improve security if you're concerned about
 people being able to see what databases other users have. See also
-@code{skip_show_database}.
+@code{skip_show_database}.  This option is deprecated as one should instead
+use the @code{SHOW DATABASES} privilege instead.

@item @code{server_id}
 The value of the @code{--server-id} option.
@ -20327,7 +20331,7 @@ Is ON if we only allow local (socket) connections.

@item @code{skip_show_database}
 This prevents people from doing @code{SHOW DATABASES} if they don't have
-the @code{PROCESS} privilege. This can improve security if you're
+the @code{SHOW DATABASE} privilege. This can improve security if you're
 concerned about people being able to see what databases other users
 have. See also @code{safe_show_database}.

@ -23506,17 +23510,21 @@ will be logged in the execution order.
 Updates to non-transactional tables are stored in the binary log
 immediately after execution.  For transactional tables such as @code{BDB}
 or @code{InnoDB} tables, all updates (@code{UPDATE}, @code{DELETE}
-or @code{INSERT}) that change tables are cached until a @code{COMMIT}.
+or @code{INSERT}) that change tables are cached until a @code{COMMIT} command
+is sent to the server. At this point mysqld writes the whole transaction to
+the binary log before the @code{COMMIT} is executed.
 Every thread will, on start, allocate a buffer of @code{binlog_cache_size}
 to buffer queries.  If a query is bigger than this, the thread will open
-a temporary file to handle the bigger cache.  The temporary file will
+a temporary file to store the transcation.  The temporary file will
 be deleted when the thread ends.

-The @code{max_binlog_cache_size} can be used to restrict the total size used
-to cache a multi-query transaction.
+The @code{max_binlog_cache_size} (default 4G) can be used to restrict
+the total size used to cache a multi-query transaction.  If a transaction is
+bigger than this it will fail and roll back.

 If you are using the update or binary log, concurrent inserts will
-not work together with @code{CREATE ... SELECT} and @code{INSERT ... SELECT}.
+be converted to normal inserts when using @code{CREATE ... SELECT} and
+@code{INSERT ... SELECT}.
 This is to ensure that you can recreate an exact copy of your tables by
 applying the log on a backup.

@ -23683,7 +23691,7 @@ started}, your slaves may fail.

 Please see the following table for an indication of master-slave
 compatibility between different versions. With regard to version 4.0,
-we recommend using same version on both sides.
+we recommend using at least 4.0.4 on both sides.

@c FIX arjen 2002-07-17 new table, not yet measured for XML/DocBook.
@multitable @columnfractions .10 .15 .15 .10 .10 .10
@ -24350,7 +24358,7 @@ may be used with @code{IO_THREAD} and @code{SQL_THREAD} options. (Slave)
 @tab Re-enables update logging if the user has the @code{SUPER} privilege.
 Ignored otherwise. (Master)

-@item @code{GLOBAL SET SQL_SLAVE_SKIP_COUNTER=n}
+@item @code{SET GLOBAL SQL_SLAVE_SKIP_COUNTER=n}
 @tab Skip the next @code{n} events from the master. Only valid when
 the slave thread is not running, otherwise, gives an error. Useful for
 recovering from replication glitches.
@ -25510,7 +25518,7 @@ temporary table to hold the result.  This typically happens if you do an
@code{ORDER BY} on a different column set than you did a @code{GROUP
 BY} on.

-@item Where used
+@item Using where (was @code{where used})
 A @code{WHERE} clause will be used to restrict which rows will be
 matched against the next table or sent to the client.  If you don't have
 this information and the table is of type @code{ALL} or @code{index},
@ -25670,7 +25678,7 @@ Executing the @code{EXPLAIN} statement again produces this result:

@example
 table type   possible_keys   key     key_len ref         rows    Extra
-tt    ALL    AssignedPC,ClientID,ActualPC NULL NULL NULL 3872    where used
+tt    ALL    AssignedPC,ClientID,ActualPC NULL NULL NULL 3872    Using where
 do    ALL    PRIMARY         NULL    NULL    NULL        2135
      range checked for each record (key map: 1)
 et_1  ALL    PRIMARY         NULL    NULL    NULL        74
@ -25696,7 +25704,7 @@ Now @code{EXPLAIN} produces the output shown here:
@example
 table type   possible_keys   key      key_len ref           rows Extra
 et    ALL    PRIMARY         NULL     NULL    NULL          74
-tt    ref    AssignedPC,     ActualPC 15      et.EMPLOYID   52   where used
+tt    ref    AssignedPC,     ActualPC 15      et.EMPLOYID   52   Using where
             ClientID,
             ActualPC
 et_1  eq_ref PRIMARY         PRIMARY  15      tt.AssignedPC 1
@ -25719,7 +25727,7 @@ Now the join is perfect, and @code{EXPLAIN} produces this result:

@example
 table type   possible_keys key     key_len ref           rows Extra
-tt    ALL    AssignedPC    NULL    NULL    NULL          3872 where used
+tt    ALL    AssignedPC    NULL    NULL    NULL          3872 Using where
             ClientID,
             ActualPC
 et    eq_ref PRIMARY       PRIMARY 15      tt.ActualPC   1
@ -28868,7 +28876,6 @@ and if you can use @code{GLOBAL} or @code{SESSION} with them.
@item read_buffer_size @tab num @tab GLOBAL | SESSION
@item read_rnd_buffer_size @tab num @tab GLOBAL | SESSION
@item rpl_recovery_rank @tab num @tab GLOBAL
-@item safe_show_database @tab bool @tab GLOBAL
@item server_id @tab num @tab GLOBAL
@item slave_compressed_protocol @tab bool @tab GLOBAL
@item slave_net_timeout @tab num @tab GLOBAL
@ -38698,8 +38705,8 @@ SUM_OVER_ALL_KEYS(max_length_of_key + sizeof(char*) * 2)
 * Table and index::             Table and Index Structures
 * File space management::       File Space Management and Disk I/O
 * Error handling::              Error Handling
-* InnoDB change history::       InnoDB Change History
 * InnoDB restrictions::         Restrictions on InnoDB Tables
+* InnoDB change history::       InnoDB Change History
 * InnoDB contact information::  InnoDB Contact Information.
@end menu

@ -50735,6 +50742,7 @@ this means that the version has not yet been released!
@c Please don't add a new version here without also updating ../configure.in!

@menu
+* News-4.1.x::                  
 * News-4.0.x::                  Changes in release 4.0.x (Beta)
 * News-3.23.x::                 Changes in release 3.23.x (Stable)
 * News-3.22.x::                 Changes in release 3.22.x (Older; Still supported)
@ -50744,7 +50752,77 @@ this means that the version has not yet been released!
@end menu


-@node News-4.0.x, News-3.23.x, News, News
+@node News-4.1.x, News-4.0.x, News, News
+@appendixsec Changes in release 4.1.x (Alpha)
+
+@cindex changes, version 4.1
+
+Version 4.1 of the MySQL server includes many enhancements and new features:
+
+@itemize @bullet
+@item
+Sub selects: @code{SELECT * from t1 where t1.a=(SELECT t2.b FROM t2)}.
+@item
+Character sets to be defined per column, table and database.
+@item
+Unicode (UTF8) support.
+@item
+Derived tables: @code{SELECT a from t1, (select * from t2) WHERE t1.a=t2.a}
+@item
+@code{BTREE} index on @code{HEAP} tables.
+@item
+Support for GIS (Geometrical data).
+@item
+@code{SHOW WARNINGS}; Shows warnings for the last command.
+@end itemize
+
+For a full list of changes, please refer to the changelog sections for
+each individual 4.1.x release.
+
+@menu
+* News-4.1.0::                  
+@end menu
+
+@node News-4.1.0,  , News-4.1.x, News-4.1.x
+@appendixsubsec Changes in release 4.1.0
+@itemize
+@item
+One can specify many temporary directories to be used in a round-robin
+fasion with:  @code{--tmpdir=dirname1:dirname2:dirname3}.
+@item
+Sub selects: @code{SELECT * from t1 where t1.a=(SELECT t2.b FROM t2)}.
+@item
+Character sets to be defined per column, table and database.
+@item
+Unicode (UTF8) support.
+@item
+Derived tables: @code{SELECT a from t1, (select * from t2) WHERE t1.a=t2.a}
+@item
+@code{BTREE} index on @code{HEAP} tables.
+@item
+Faster embedded server.
+@item
+One can add a comment per column in @code{CREATE TABLE}.
+@item
+@code{SHOW FULL COLUMNS FROM table_name} shows column comments.
+@item
+@code{ALTER DATABASE}.
+@item
+Support for GIS (Geometrical data).
+@item
+@code{SHOW WARNINGS}; Shows warnings from the last command.
+@item
+One can specify a column type for a colum in @code{CREATE TABLE
+... SELECT} by defining the column in the @code{CREATE} part
+
+@example
+CREATE TABLE foo (a tinyint not null) SELECT b+1 AS 'a' FROM bar;
+@end example
+
+@end itemize
+
+
+@node News-4.0.x, News-3.23.x, News-4.1.x, News
@appendixsec Changes in release 4.0.x (Beta)

@cindex changes, version 4.0
@ -50815,6 +50893,19 @@ each individual 4.0.x release.
@appendixsubsec Changes in release 4.0.5
@itemize
@item
+When one uses the @code{--open-files-limit=#} option to @code{mysqld_safe}
+it's now passed on to @code{mysqld}
+@item
+Fixed that @code{GROUP BY} on columns that may have a @code{NULL} value
+doesn't always use disk based temporary tables.
+@item
+Changed output from @code{EXPLAIN} from @code{'where used'} to
+@code{'Using where'} to make it more in line with other output.
+@item
+Removed variable @code{safe_show_database} as it was not used anymore.
+@item
+Read @code{--des-key-file} relative to database directory.
+@item
 Small code improvement in multi-table updates
@item
 Fixed a newly introduced bug that caused @code{ORDER BY ... LIMIT #}
@ -51680,6 +51771,11 @@ not yet 100% confident in this code.
@appendixsubsec Changes in release 3.23.54
@itemize
@item
+Allow one to start multiple MySQL servers on windows (code backported
+from 4.0.2).
+@item
+Fixed that @code{--core-file} works on Linux (at least on kernel 2.4.18).
+@item
 Fixed a problem with BDB and @code{ALTER TABLE}.
@item
 Fixed reference to freed memory when doing complicated @code{GROUP BY
@ -51765,6 +51861,11 @@ Changed initialisation of @code{RND()} to make it less predicatable.
 Fixed problem with @code{GROUP BY} on result with expression that created a
@code{BLOB} field.
@item
+Fixed problem with @code{GROUP BY} on columns that have NULL values.  To
+solve this we now create an MyISAM temporary table when doing a group by
+on a possible NULL item.  In MySQL 4.0.5 we can again use in memory HEAP
+tables for this case.
+@item
 Fixed problem with privilege tables when downgrading from 4.0.2 to 3.23.
@item
 Fixed thread bug in @code{SLAVE START}, @code{SLAVE STOP} and automatic repair
@ -52506,8 +52607,9 @@ long as @code{server-id} is set and valid @file{master.info} is present.
 Partial updates (terminated with kill) are now logged with a special error
 code to the binary log. Slave will refuse to execute them if the error code
 indicates the update was terminated abnormally, and will have to be recovered
-with @code{SET SQL_SLAVE_SKIP_COUNTER=1; SLAVE START} after a manual sanity
-check/correction of data integrity.
+with @code{SET SQL_SLAVE_SKIP_COUNTER=1; SLAVE START} after a manual
+sanity check/correction of data integrity.  Update: In 4.0.3 and above
+you have to use @code{SET GLOBAL}.
@item
 Fixed bug that erroneously logged a drop of internal temporary table
 on thread termination to the binary log -- this bug affected replication.
--- a/acconfig.h
+++ b/acconfig.h
@ -274,7 +274,7 @@
 #undef SPRINTF_RETURNS_INT
 #undef SPRINTF_RETURNS_GARBAGE

-/* Needed to get large file supportat HPUX 10.20 */
+/* Needed to get large file support on HPUX 10.20 */
 #undef __STDC_EXT__

 #undef STACK_DIRECTION
--- a/acinclude.m4
+++ b/acinclude.m4
@ -1190,5 +1190,142 @@ dnl ---------------------------------------------------------------------------
 dnl END OF MYSQL_CHECK_INNODB SECTION
 dnl ---------------------------------------------------------------------------

+dnl By default, many hosts won't let programs access large files;
+dnl one must use special compiler options to get large-file access to work.
+dnl For more details about this brain damage please see:
+dnl http://www.sas.com/standards/large.file/x_open.20Mar96.html
+
+dnl Written by Paul Eggert <eggert@twinsun.com>.
+
+dnl Internal subroutine of AC_SYS_LARGEFILE.
+dnl AC_SYS_LARGEFILE_FLAGS(FLAGSNAME)
+AC_DEFUN(AC_SYS_LARGEFILE_FLAGS,
+  [AC_CACHE_CHECK([for $1 value to request large file support],
+     ac_cv_sys_largefile_$1,
+     [if ($GETCONF LFS_$1) >conftest.1 2>conftest.2 && test ! -s conftest.2
+      then
+        ac_cv_sys_largefile_$1=`cat conftest.1`
+      else
+	ac_cv_sys_largefile_$1=no
+	ifelse($1, CFLAGS,
+	  [case "$host_os" in
+	   # HP-UX 10.20 requires -D__STDC_EXT__ with gcc 2.95.1.
+changequote(, )dnl
+	   hpux10.[2-9][0-9]* | hpux1[1-9]* | hpux[2-9][0-9]*)
+changequote([, ])dnl
+	     if test "$GCC" = yes; then
+	       ac_cv_sys_largefile_CFLAGS=-D__STDC_EXT__
+	     fi
+	     ;;
+	   # IRIX 6.2 and later require cc -n32.
+changequote(, )dnl
+	   irix6.[2-9]* | irix6.1[0-9]* | irix[7-9].* | irix[1-9][0-9]*)
+changequote([, ])dnl
+	     if test "$GCC" != yes; then
+	       ac_cv_sys_largefile_CFLAGS=-n32
+	     fi
+	   esac
+	   if test "$ac_cv_sys_largefile_CFLAGS" != no; then
+	     ac_save_CC="$CC"
+	     CC="$CC $ac_cv_sys_largefile_CFLAGS"
+	     AC_TRY_LINK(, , , ac_cv_sys_largefile_CFLAGS=no)
+	     CC="$ac_save_CC"
+	   fi])
+      fi
+      rm -f conftest*])])
+
+dnl Internal subroutine of AC_SYS_LARGEFILE.
+dnl AC_SYS_LARGEFILE_SPACE_APPEND(VAR, VAL)
+AC_DEFUN(AC_SYS_LARGEFILE_SPACE_APPEND,
+  [case $2 in
+   no) ;;
+   ?*)
+     case "[$]$1" in
+     '') $1=$2 ;;
+     *) $1=[$]$1' '$2 ;;
+     esac ;;
+   esac])
+
+dnl Internal subroutine of AC_SYS_LARGEFILE.
+dnl AC_SYS_LARGEFILE_MACRO_VALUE(C-MACRO, CACHE-VAR, COMMENT, CODE-TO-SET-DEFAULT)
+AC_DEFUN(AC_SYS_LARGEFILE_MACRO_VALUE,
+  [AC_CACHE_CHECK([for $1], $2,
+     [$2=no
+changequote(, )dnl
+      for ac_flag in $ac_cv_sys_largefile_CFLAGS no; do
+	case "$ac_flag" in
+	-D$1)
+	  $2=1 ;;
+	-D$1=*)
+	  $2=`expr " $ac_flag" : '[^=]*=\(.*\)'` ;;
+	esac
+      done
+      $4
+changequote([, ])dnl
+      ])
+   if test "[$]$2" != no; then
+     AC_DEFINE_UNQUOTED([$1], [$]$2, [$3])
+   fi])
+
+AC_DEFUN(MYSQL_SYS_LARGEFILE,
+  [AC_REQUIRE([AC_CANONICAL_HOST])
+  AC_ARG_ENABLE(largefile,
+     [  --disable-largefile     Omit support for large files])
+   if test "$enable_largefile" != no; then
+     AC_CHECK_TOOL(GETCONF, getconf)
+     AC_SYS_LARGEFILE_FLAGS(CFLAGS)
+     AC_SYS_LARGEFILE_FLAGS(LDFLAGS)
+     AC_SYS_LARGEFILE_FLAGS(LIBS)
+
+     for ac_flag in $ac_cv_sys_largefile_CFLAGS no; do
+       case "$ac_flag" in
+       no) ;;
+       -D_FILE_OFFSET_BITS=*) ;;
+       -D_LARGEFILE_SOURCE | -D_LARGEFILE_SOURCE=*) ;;
+       -D_LARGE_FILES | -D_LARGE_FILES=*) ;;
+       -D?* | -I?*)
+	 AC_SYS_LARGEFILE_SPACE_APPEND(CPPFLAGS, "$ac_flag") ;;
+       *)
+	 AC_SYS_LARGEFILE_SPACE_APPEND(CFLAGS, "$ac_flag") ;;
+       esac
+     done
+     AC_SYS_LARGEFILE_SPACE_APPEND(LDFLAGS, "$ac_cv_sys_largefile_LDFLAGS")
+     AC_SYS_LARGEFILE_SPACE_APPEND(LIBS, "$ac_cv_sys_largefile_LIBS")
+
+     AC_SYS_LARGEFILE_MACRO_VALUE(_FILE_OFFSET_BITS,
+       ac_cv_sys_file_offset_bits,
+       [Number of bits in a file offset, on hosts where this is settable.],
+       [case "$host_os" in
+	# HP-UX 10.20 and later
+	hpux10.[2-9][0-9]* | hpux1[1-9]* | hpux[2-9][0-9]*)
+	  ac_cv_sys_file_offset_bits=64 ;;
+	# We can't declare _FILE_OFFSET_BITS here as this will cause
+	# compile errors as AC_PROG_CC adds include files in confdefs.h
+	# We solve this (until autoconf is fixed) by instead declaring it
+	# as define instead
+	solaris2.[8,9])
+	  CFLAGS="$CFLAGS -D_FILE_OFFSET_BITS=64"
+	  CXXFLAGS="$CXXFLAGS -D_FILE_OFFSET_BITS=64"
+	  ac_cv_sys_file_offset_bits=no ;;
+	esac])
+     AC_SYS_LARGEFILE_MACRO_VALUE(_LARGEFILE_SOURCE,
+       ac_cv_sys_largefile_source,
+       [Define to make fseeko etc. visible, on some hosts.],
+       [case "$host_os" in
+	# HP-UX 10.20 and later
+	hpux10.[2-9][0-9]* | hpux1[1-9]* | hpux[2-9][0-9]*)
+	  ac_cv_sys_largefile_source=1 ;;
+	esac])
+     AC_SYS_LARGEFILE_MACRO_VALUE(_LARGE_FILES,
+       ac_cv_sys_large_files,
+       [Define for large files, on AIX-style hosts.],
+       [case "$host_os" in
+	# AIX 4.2 and later
+	aix4.[2-9]* | aix4.1[0-9]* | aix[5-9].* | aix[1-9][0-9]*)
+	  ac_cv_sys_large_files=1 ;;
+	esac])
+   fi
+  ])
+
 dnl ---------------------------------------------------------------------------

--- a/client/client_priv.h
+++ b/client/client_priv.h
@ -38,4 +38,4 @@ enum options { OPT_CHARSETS_DIR=256, OPT_DEFAULT_CHARSET,
               OPT_SSL_KEY, OPT_SSL_CERT, OPT_SSL_CA, OPT_SSL_CAPATH,
               OPT_SSL_CIPHER, OPT_SHUTDOWN_TIMEOUT, OPT_LOCAL_INFILE,
               OPT_PROMPT, OPT_IGN_LINES,OPT_TRANSACTION,OPT_MYSQL_PROTOCOL,
-               OPT_SHARED_MEMORY_BASE_NAME };
+               OPT_SHARED_MEMORY_BASE_NAME, OPT_FRM };
--- a/client/mysqlbinlog.cc
+++ b/client/mysqlbinlog.cc
@ -176,7 +176,7 @@ static void dump_remote_file(NET* net, const char* fname)
 }


-static my_bool
+extern "C" my_bool
 get_one_option(int optid, const struct my_option *opt __attribute__((unused)),
 	       char *argument)
 {
@ -210,7 +210,6 @@ get_one_option(int optid, const struct my_option *opt __attribute__((unused)),
    print_version();
    exit(0);
  case '?':
-  default:
    usage();
    exit(0);
  }
--- a/client/mysqlcheck.c
+++ b/client/mysqlcheck.c
@ -34,7 +34,7 @@ static my_bool opt_alldbs = 0, opt_check_only_changed = 0, opt_extended = 0,
               opt_compress = 0, opt_databases = 0, opt_fast = 0,
               opt_medium_check = 0, opt_quick = 0, opt_all_in_1 = 0,
               opt_silent = 0, opt_auto_repair = 0, ignore_errors = 0,
-               tty_password = 0;
+               tty_password = 0, opt_frm = 0;
 static uint verbose = 0, opt_mysql_port=0;
 static my_string opt_mysql_unix_port = 0;
 static char *opt_password = 0, *current_user = 0, *default_charset = 0,
@ -139,6 +139,10 @@ static struct my_option my_long_options[] =
  {"user", 'u', "User for login if not current user.", (gptr*) &current_user,
   (gptr*) &current_user, 0, GET_STR, REQUIRED_ARG, 0, 0, 0, 0, 0, 0},
 #endif
+  {"use-frm", OPT_FRM,
+   "When used with REPAIR, get table structure from .frm file, so the table can be repaired even if .MYI header is corrupted.",
+   (gptr*) &opt_frm, (gptr*) &opt_frm, 0, GET_BOOL, NO_ARG, 0, 0, 0, 0, 0,
+   0},
  {"verbose", 'v', "Print info about the various stages.", 0, 0, 0, GET_NO_ARG,
   NO_ARG, 0, 0, 0, 0, 0, 0},
  {"version", 'V', "Output version information and exit.", 0, 0, 0, GET_NO_ARG,
@ -368,21 +372,25 @@ static int process_selected_tables(char *db, char **table_names, int tables)
    return 1;
  if (opt_all_in_1)
  {
+    /* 
+      We need table list in form `a`, `b`, `c`
+      that's why we need 4 more chars added to to each table name
+      space is for more readable output in logs and in case of error
+    */	  
    char *table_names_comma_sep, *end;
    int i, tot_length = 0;

    for (i = 0; i < tables; i++)
-      tot_length += strlen(*(table_names + i)) + 1;
+      tot_length += strlen(*(table_names + i)) + 4;

    if (!(table_names_comma_sep = (char *)
-	  my_malloc((sizeof(char) * tot_length) + 1, MYF(MY_WME))))
+	  my_malloc((sizeof(char) * tot_length) + 4, MYF(MY_WME))))
      return 1;

    for (end = table_names_comma_sep + 1; tables > 0;
 	 tables--, table_names++)
    {
-      end = strmov(end, *table_names);
-      *end++= ',';
+      end = strxmov(end, " `", *table_names, "`,", NullS);
    }
    *--end = 0;
    handle_request_for_tables(table_names_comma_sep + 1, tot_length - 1);
@ -409,22 +417,27 @@ static int process_all_tables_in_db(char *database)

  if (opt_all_in_1)
  {
+    /* 
+      We need table list in form `a`, `b`, `c`
+      that's why we need 4 more chars added to to each table name
+      space is for more readable output in logs and in case of error
+     */
+	  
    char *tables, *end;
    uint tot_length = 0;

    while ((row = mysql_fetch_row(res)))
-      tot_length += strlen(row[0]) + 1;
+      tot_length += strlen(row[0]) + 4;
    mysql_data_seek(res, 0);

-    if (!(tables=(char *) my_malloc(sizeof(char)*tot_length+1, MYF(MY_WME))))
+    if (!(tables=(char *) my_malloc(sizeof(char)*tot_length+4, MYF(MY_WME))))
    {
      mysql_free_result(res);
      return 1;
    }
    for (end = tables + 1; (row = mysql_fetch_row(res)) ;)
    {
-      end = strmov(end, row[0]);
-      *end++= ',';
+      end = strxmov(end, " `", row[0], "`,", NullS);	    
    }
    *--end = 0;
    if (tot_length)
@ -472,6 +485,7 @@ static int handle_request_for_tables(char *tables, uint length)
    op = "REPAIR";
    if (opt_quick)              end = strmov(end, " QUICK");
    if (opt_extended)           end = strmov(end, " EXTENDED");
+    if (opt_frm)                end = strmov(end, " USE_FRM");
    break;
  case DO_ANALYZE:
    op = "ANALYZE";
@ -483,10 +497,14 @@ static int handle_request_for_tables(char *tables, uint length)

  if (!(query =(char *) my_malloc((sizeof(char)*(length+110)), MYF(MY_WME))))
    return 1;
+  if (opt_all_in_1)
+    /* No backticks here as we added them before */
    sprintf(query, "%s TABLE %s %s", op, tables, options);
+  else
+    sprintf(query, "%s TABLE `%s` %s", op, tables, options);
  if (mysql_query(sock, query))
  {
-    sprintf(message, "when executing '%s TABLE `%s` %s", op, tables,options);
+    sprintf(message, "when executing '%s TABLE ... %s'", op, options);
    DBerror(sock, message);
    return 1;
  }
--- a/client/mysqldump.c
+++ b/client/mysqldump.c
@ -305,7 +305,10 @@ static void short_usage(void)
 static void write_header(FILE *sql_file, char *db_name)
 {
  if (opt_xml)
+  {
    fprintf(sql_file,"<?xml version=\"1.0\"?>\n");
+    fprintf(sql_file,"<mysqldump>\n");
+  }
  else
  {
    fprintf(sql_file, "-- MySQL dump %s\n--\n", DUMP_VERSION);
@ -319,6 +322,12 @@ static void write_header(FILE *sql_file, char *db_name)
  return;
 } /* write_header */

+static void write_footer(FILE *sql_file)
+{
+  if (opt_xml)
+    fprintf(sql_file,"</mysqldump>");
+  fputs("\n", sql_file);
+} /* write_footer */

 static my_bool
 get_one_option(int optid, const struct my_option *opt __attribute__((unused)),
@ -1004,7 +1013,7 @@ static void dumpTable(uint numFields, char *table)
    rownr=0;
    init_length=(uint) strlen(insert_pat)+4;
    if (opt_xml)
-      fprintf(md_result_file, "\t<%s>\n", table);
+      fprintf(md_result_file, "\t<table name=\"%s\">\n", table);

    if (opt_autocommit)
      fprintf(md_result_file, "set autocommit=0;\n");
@ -1094,9 +1103,9 @@ static void dumpTable(uint numFields, char *table)
 	      /* change any strings ("inf","nan",..) into NULL */
 	      char *ptr = row[i];
 	      if (opt_xml)
-		fprintf(md_result_file, "\t\t<%s>%s</%s>\n",
+		fprintf(md_result_file, "\t\t<field name=\"%s\">%s</field>\n",
 			field->name,
-			!my_isalpha(system_charset_info,*ptr) ?ptr: "NULL",field->name);
+			!my_isalpha(system_charset_info, *ptr) ? ptr: "NULL");
 	      else
 		fputs((!my_isalpha(system_charset_info,*ptr)) ? 
 		       ptr : "NULL", md_result_file);
@ -1105,8 +1114,8 @@ static void dumpTable(uint numFields, char *table)
 	  else
 	  {
 	    if (opt_xml)
-	      fprintf(md_result_file, "\t\t<%s>%s</%s>\n",
-		      field->name, "NULL", field->name);
+	      fprintf(md_result_file, "\t\t<field name=\"%s\">%s</field>\n",
+		      field->name, "NULL");
 	    else
 	      fputs("NULL", md_result_file);
 	  }
@ -1147,7 +1156,7 @@ static void dumpTable(uint numFields, char *table)

    /* XML - close table tag and supress regular output */
    if (opt_xml)
-	fprintf(md_result_file, "\t</%s>\n", table);
+	fprintf(md_result_file, "\t</table>\n");
    else if (extended_insert && row_break)
      fputs(";\n", md_result_file);		/* If not empty table */
    fflush(md_result_file);
@ -1179,7 +1188,7 @@ static void print_quoted_xml(FILE *output, char *fname, char *str, uint len)
 {
  const char *end;

-  fprintf(output, "\t\t<%s>", fname);
+  fprintf(output, "\t\t<field name=\"%s\">", fname);
  for (end = str + len; str != end; str++)
  {
    if (*str == '<')
@ -1193,7 +1202,7 @@ static void print_quoted_xml(FILE *output, char *fname, char *str, uint len)
    else
      fputc(*str, output);
  }
-  fprintf(output, "</%s>\n", fname);
+  fprintf(output, "</field>\n");
 }

 static char *getTableName(int reset)
@ -1248,13 +1257,8 @@ static int dump_databases(char **db_names)
  int result=0;
  for ( ; *db_names ; db_names++)
  {
-    /* XML edit - add database element */
-    if (opt_xml)
-      fprintf(md_result_file, "<%s>\n", *db_names);
    if (dump_all_tables_in_db(*db_names))
      result=1;
-    if (opt_xml)
-      fprintf(md_result_file, "</%s>\n", *db_names);
  }
  return result;
 } /* dump_databases */
@ -1267,7 +1271,7 @@ static int init_dumping(char *database)
    DBerror(sock, "when selecting the database");
    return 1;			/* If --force */
  }
-  if (!path)
+  if (!path && !opt_xml)
  {
    if (opt_databases || opt_alldbs)
    {
@ -1313,6 +1317,8 @@ static int dump_all_tables_in_db(char *database)

  if (init_dumping(database))
    return 1;
+  if (opt_xml)
+    fprintf(md_result_file, "<database name=\"%s\">\n", database);
  if (lock_tables)
  {
    DYNAMIC_STRING query;
@ -1339,6 +1345,8 @@ static int dump_all_tables_in_db(char *database)
    if (!dFlag && numrows > 0)
      dumpTable(numrows,table);
  }
+  if (opt_xml)
+    fprintf(md_result_file, "</database>\n");
  if (lock_tables)
    mysql_query(sock,"UNLOCK_TABLES");
  return 0;
@ -1375,12 +1383,16 @@ static int dump_selected_tables(char *db, char **table_names, int tables)
      DBerror(sock, "when doing refresh");
     /* We shall countinue here, if --force was given */
  }
+  if (opt_xml)
+    fprintf(md_result_file, "<database name=\"%s\">\n", db);
  for (; tables > 0 ; tables-- , table_names++)
  {
    numrows = getTableStructure(*table_names, db);
    if (!dFlag && numrows > 0)
      dumpTable(numrows, *table_names);
  }
+  if (opt_xml)
+    fprintf(md_result_file, "</database>\n");
  if (lock_tables)
    mysql_query(sock,"UNLOCK_TABLES");
  return 0;
@ -1514,7 +1526,7 @@ int main(int argc, char **argv)
    }		      
  }
  dbDisconnect(current_host);
-  fputs("\n", md_result_file);
+  write_footer(md_result_file);
  if (md_result_file != stdout)
    my_fclose(md_result_file, MYF(0));
  my_free(opt_password, MYF(MY_ALLOW_ZERO_PTR));
--- a/client/mysqlshow.c
+++ b/client/mysqlshow.c
@ -56,6 +56,7 @@ static my_string opt_mysql_unix_port=0;
 int main(int argc, char **argv)
 {
  int error;
+  my_bool first_argument_uses_wildcards=0;
  char *wild;
  MYSQL mysql;
  MY_INIT(argv[0]);
@ -63,21 +64,37 @@ int main(int argc, char **argv)
  get_options(&argc,&argv);

  wild=0;
-  if (argc && strcont(argv[argc-1],"*?%_"))
+  if (argc)
  {
-    char *pos;
-
-    wild=argv[--argc];
-    for (pos=wild ; *pos ; pos++)
-    {					/* Unix wildcards to sql  */
-      if (*pos == '*')
-	*pos='%';
-      else if (*pos == '?')
-	*pos='_';
+    char *pos= argv[argc-1], *to;
+    for (to= pos ; *pos ; pos++, to++)
+    {
+      switch (*pos)
+      {
+      case '*':
+	*pos= '%';
+	first_argument_uses_wildcards= 1;
+	break;
+      case '?':
+	*pos= '_';
+	first_argument_uses_wildcards= 1;
+	break;
+      case '%':
+      case '_':
+	first_argument_uses_wildcards= 1;
+	break;
+      case '\\':
+	pos++;
+      default: break;
      }
+      *to= *pos;
    }    
+    *to= *pos; // just to copy a '\0'  if '\\' was used
+  }
+  if (first_argument_uses_wildcards)
+    wild= argv[--argc];
  else if (argc == 3)			/* We only want one field */
-    wild=argv[--argc];
+    wild= argv[--argc];

  if (argc > 2)
  {
@ -99,7 +116,7 @@ int main(int argc, char **argv)
    mysql_options(&mysql,MYSQL_SHARED_MEMORY_BASE_NAME,shared_memory_base_name);
 #endif
  if (!(mysql_real_connect(&mysql,host,user,opt_password,
-			   argv[0],opt_mysql_port,opt_mysql_unix_port,
+			   (first_argument_uses_wildcards) ? "" : argv[0],opt_mysql_port,opt_mysql_unix_port,
 			   0)))
  {
    fprintf(stderr,"%s: %s\n",my_progname,mysql_error(&mysql));
--- a/client/mysqltest.c
+++ b/client/mysqltest.c
@ -91,7 +91,7 @@


 enum {OPT_MANAGER_USER=256,OPT_MANAGER_HOST,OPT_MANAGER_PASSWD,
-      OPT_MANAGER_PORT,OPT_MANAGER_WAIT_TIMEOUT};
+      OPT_MANAGER_PORT,OPT_MANAGER_WAIT_TIMEOUT, OPT_SKIP_SAFEMALLOC};

 static int record = 0, opt_sleep=0;
 static char *db = 0, *pass=0;
@ -1850,6 +1850,9 @@ static struct my_option my_long_options[] =
   0, 0, 0, GET_STR, REQUIRED_ARG, 0, 0, 0, 0, 0, 0},
  {"silent", 's', "Suppress all normal output. Synonym for --quiet.",
   (gptr*) &silent, (gptr*) &silent, 0, GET_BOOL, NO_ARG, 0, 0, 0, 0, 0, 0},
+  {"skip-safemalloc", OPT_SKIP_SAFEMALLOC,
+   "Don't use the memory allocation checking", 0, 0, 0, GET_NO_ARG, NO_ARG,
+   0, 0, 0, 0, 0, 0},
  {"sleep", 'T', "Sleep always this many seconds on sleep commands",
   (gptr*) &opt_sleep, (gptr*) &opt_sleep, 0, GET_INT, REQUIRED_ARG, 0, 0, 0,
   0, 0, 0},
@ -1949,6 +1952,11 @@ get_one_option(int optid, const struct my_option *opt __attribute__((unused)),
    if (read_server_arguments(argument))
      die(NullS);
    break;
+  case OPT_SKIP_SAFEMALLOC:
+#ifdef SAFEMALLOC
+    sf_malloc_quick=1;
+#endif
+    break;
  case 'V':
    print_version();
    exit(0);
@ -2323,6 +2331,7 @@ static void var_from_env(const char* name, const char* def_val)
 static void init_var_hash()
 {
  VAR* v;
+  DBUG_ENTER("init_var_hash");
  if (hash_init(&var_hash, system_charset_info, 
                1024, 0, 0, get_var_key, var_free, MYF(0)))
    die("Variable hash initialization failed");
@ -2332,6 +2341,7 @@ static void init_var_hash()
  var_from_env("BIG_TEST", opt_big_test ? "1" : "0");
  v=var_init(0,"MAX_TABLES", 0, (sizeof(ulong) == 4) ? "31" : "63",0);
  hash_insert(&var_hash, (byte*)v);
+  DBUG_VOID_RETURN;
 }


--- a/configure.in
+++ b/configure.in
@ -72,12 +72,14 @@ case $MACHINE_TYPE in
 esac

 # Save some variables and the command line options for mysqlbug
+SAVE_ASFLAGS="$ASFLAGS"
 SAVE_CFLAGS="$CFLAGS"
 SAVE_CXXFLAGS="$CXXFLAGS"
 SAVE_LDFLAGS="$LDFLAGS"
 SAVE_CXXLDFLAGS="$CXXLDFLAGS"
 CONF_COMMAND="$0 $ac_configure_args"
 AC_SUBST(CONF_COMMAND)
+AC_SUBST(SAVE_ASFLAGS)
 AC_SUBST(SAVE_CFLAGS)
 AC_SUBST(SAVE_CXXFLAGS)
 AC_SUBST(SAVE_LDFLAGS)
@ -602,8 +604,9 @@ AC_ARG_ENABLE(assembler,
 AC_MSG_CHECKING(if we should use assembler functions)
 # For now we only support assembler on i386 and sparc systems
 AM_CONDITIONAL(ASSEMBLER_x86, test "$ENABLE_ASSEMBLER" = "yes" -a "$BASE_MACHINE_TYPE" = "i386")
-AM_CONDITIONAL(ASSEMBLER_sparc, test "$ENABLE_ASSEMBLER" = "yes" -a "$BASE_MACHINE_TYPE" = "sparc")
-AM_CONDITIONAL(ASSEMBLER, test "$ASSEMBLER_x86_TRUE" = "" -o "$ASSEMBLER_sparc_TRUE" = "")
+AM_CONDITIONAL(ASSEMBLER_sparc32, test "$ENABLE_ASSEMBLER" = "yes" -a "$BASE_MACHINE_TYPE" = "sparc")
+AM_CONDITIONAL(ASSEMBLER_sparc64, test "$ENABLE_ASSEMBLER" = "yes" -a "$BASE_MACHINE_TYPE" = "sparcv9")
+AM_CONDITIONAL(ASSEMBLER, test "$ASSEMBLER_x86_TRUE" = "" -o "$ASSEMBLER_sparc32_TRUE" = "")

 if test "$ASSEMBLER_TRUE" = ""
 then
@ -674,7 +677,7 @@ else
  AC_MSG_RESULT([no])
 fi

-AC_SYS_LARGEFILE
+MYSQL_SYS_LARGEFILE

 # Types that must be checked AFTER large file support is checked
 AC_TYPE_SIZE_T
@ -917,8 +920,8 @@ case $SYSTEM_TYPE in
    ;;
  *hpux10.20*)
    echo "Enabling workarounds for hpux 10.20"
-    CFLAGS="$CFLAGS -DHAVE_BROKEN_SNPRINTF -DSIGNALS_DONT_BREAK_READ -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHPUX -DSIGNAL_WITH_VIO_CLOSE -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT -DHAVE_POSIX1003_4a_MUTEX"
-    CXXFLAGS="$CXXFLAGS -DHAVE_BROKEN_SNPRINTF -D_INCLUDE_LONGLONG -DSIGNALS_DONT_BREAK_READ -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHPUX -DSIGNAL_WITH_VIO_CLOSE -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT -DHAVE_POSIX1003_4a_MUTEX"
+    CFLAGS="$CFLAGS -DHAVE_BROKEN_SNPRINTF -DSIGNALS_DONT_BREAK_READ -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHPUX10 -DSIGNAL_WITH_VIO_CLOSE -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT -DHAVE_POSIX1003_4a_MUTEX"
+    CXXFLAGS="$CXXFLAGS -DHAVE_BROKEN_SNPRINTF -D_INCLUDE_LONGLONG -DSIGNALS_DONT_BREAK_READ -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHPUX10 -DSIGNAL_WITH_VIO_CLOSE -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT -DHAVE_POSIX1003_4a_MUTEX"
    if test "$with_named_thread" = "no"
    then 
      echo "Using --with-named-thread=-lpthread"
@ -927,13 +930,19 @@ case $SYSTEM_TYPE in
    ;;
  *hpux11.*)
    echo "Enabling workarounds for hpux 11"
-    CFLAGS="$CFLAGS -DHAVE_BROKEN_PREAD -DDONT_USE_FINITE -DHAVE_BROKEN_GETPASS -DNO_FCNTL_NONBLOCK -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT"
-    CXXFLAGS="$CXXFLAGS -DHAVE_BROKEN_PREAD -DDONT_USE_FINITE -D_INCLUDE_LONGLONG -DNO_FCNTL_NONBLOCK -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT"
+    CFLAGS="$CFLAGS -DHPUX11 -DHAVE_BROKEN_PREAD -DDONT_USE_FINITE -DHAVE_BROKEN_GETPASS -DNO_FCNTL_NONBLOCK -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT"
+    CXXFLAGS="$CXXFLAGS -DHPUX11 -DHAVE_BROKEN_PREAD -DDONT_USE_FINITE -D_INCLUDE_LONGLONG -DNO_FCNTL_NONBLOCK -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT"
    if test "$with_named_thread" = "no"
    then 
      echo "Using --with-named-thread=-lpthread"
      with_named_thread="-lpthread"
    fi
+    # Fixes for HPUX 11.0 compiler
+    if test "$ac_cv_prog_gcc" = "no"
+    then
+      CFLAGS="$CFLAGS +DD64 -DHAVE_BROKEN_INLINE"
+      CXXFLAGS="$CXXFLAGS +DD64 +O2"
+    fi
    ;;
  *rhapsody*)
    if test "$ac_cv_prog_gcc" = "yes"
@ -966,7 +975,8 @@ case $SYSTEM_TYPE in
    ;;
  *freebsd*)
    echo "Adding fix for interrupted reads"
-    CXXFLAGS="$CXXFLAGS -DMYSQLD_NET_RETRY_COUNT=1000000"
+    CFLAGS="$CFLAGS -DHAVE_BROKEN_REALPATH"
+    CXXFLAGS="$CXXFLAGS -DMYSQLD_NET_RETRY_COUNT=1000000 -DHAVE_BROKEN_REALPATH"
    ;;
  *netbsd*)
    echo "Adding flag -Dunix"
--- a/dbug/dbug_add_tags.pl
+++ b/dbug/dbug_add_tags.pl
@ -0,0 +1,73 @@
+#!/usr/bin/perl
+
+die "No files specified\n" unless $ARGV[0];
+
+$ctags="exctags -x -f - --c-types=f -u";
+
+sub get_tag {
+  local $.; local $_=<TAGS>;
+  ($symbol, $line)= /^(.*\S)\s+function\s+(\d+)/;
+  $symbol=$1 if /\s(\S+)\s*\(/;
+  $line=1e50 unless $line;
+}
+
+while($src=shift)
+{
+  warn "==> $src\n";
+ 
+  $dst=$src.$$;
+  open(TAGS, "$ctags $src|") || die "Cannot exec('$ctags $src'): $!";
+  open(SRC, "<$src") || die "Cannot open $src: $!";
+  open(DST, ">$dst") || die "Cannot create $dst: $!";
+  select DST;
+
+  &get_tag;
+  $in_func=0;
+  while(<SRC>)
+  {
+    my $orig=$_;
+    if ($in_func)
+    {
+      if (/\breturn\b/ && !/\/\*.*\breturn\b.*\*\// && !/;/ )
+      {
+        $_.=<SRC> until /;/;
+      }
+      s/(?<=\s)return\s*;/DBUG_VOID_RETURN;/;
+      s/(?<=\s)return\s*(.+)\s*;/DBUG_RETURN(\1);/s;
+      $ret_line=$. if /DBUG_(VOID_)?RETURN/; #{{
+      print "$tab  DBUG_VOID_RETURN;\n" if /^$tab}/ && $ret_line < $.-1;
+      $in_func=0 if /^$tab}/;
+      warn "$src:".($.-1)."\t$orig" if /\breturn\b/;
+    }
+    print;
+    next if $. < $line;
+    die "Something wrong: \$.=$., \$line=$line, \$symbol=$symbol\n" if $. > $line;
+    &get_tag && next if /^\s*inline /;
+    print $_=<SRC> until /{/; $tab=$`;
+    &get_tag && next if /}/; # skip one-liners
+    $semicolon=1;
+    while(<SRC>)
+    {
+      $skip=!$semicolon;
+      $semicolon= /;\s*$/;
+      print && next if $skip ||
+        (/^\s+\w+((::\w+)?|<\w+>)\s+\**\w+/ && !/^\s*return/);
+      last if /DBUG_ENTER/;
+      print "$tab  DBUG_ENTER(\"$symbol\");\n";
+      print "\n" unless $_ eq "\n";
+      last;
+    }
+    $in_func=1;
+    &get_tag;
+    redo;
+  }
+  close SRC;
+  close DST;
+  close TAGS;
+  unlink("$src.orig");
+  rename($src, "$src.orig") || die "Cannot rename $src to $src.orig: $!";
+  rename($dst, $src) || die "Cannot rename $dst to $src: $!";
+}
+
+warn "All done!\n";
+
--- a/extra/Makefile.am
+++ b/extra/Makefile.am
@ -18,7 +18,7 @@ INCLUDES =		@MT_INCLUDES@ -I$(srcdir)/../include -I../include -I..
 LDADD =			@CLIENT_EXTRA_LDFLAGS@ ../mysys/libmysys.a \
 			../dbug/libdbug.a ../strings/libmystrings.a
 bin_PROGRAMS =		replace comp_err perror resolveip my_print_defaults \
-resolve_stack_dump mysql_install
+			resolve_stack_dump mysql_install

 # Don't update the files from bitkeeper
 %::SCCS/s.%
--- a/include/errmsg.h
+++ b/include/errmsg.h
@ -61,26 +61,27 @@ extern const char *client_errors[];	/* Error messages */
 #define CR_PROBE_SLAVE_HOSTS    2023
 #define CR_PROBE_SLAVE_CONNECT  2024
 #define CR_PROBE_MASTER_CONNECT 2025
+#define CR_SSL_CONNECTION_ERROR 2026

 /* new 4.1 error codes */
-#define CR_INVALID_CONN_HANDLE	2026
-#define CR_NULL_POINTER		2027
-#define CR_NO_PREPARE_STMT	2028
-#define CR_NOT_ALL_PARAMS_BOUND	2029
-#define CR_DATA_TRUNCATED	2030
-#define CR_NO_PARAMETERS_EXISTS 2031
-#define CR_INVALID_PARAMETER_NO 2032
-#define CR_INVALID_BUFFER_USE	2033
-#define CR_UNSUPPORTED_PARAM_TYPE 2034
+#define CR_INVALID_CONN_HANDLE	2027
+#define CR_NULL_POINTER		2028
+#define CR_NO_PREPARE_STMT	2029
+#define CR_NOT_ALL_PARAMS_BOUND	2030
+#define CR_DATA_TRUNCATED	2031
+#define CR_NO_PARAMETERS_EXISTS 2032
+#define CR_INVALID_PARAMETER_NO 2033
+#define CR_INVALID_BUFFER_USE	2034
+#define CR_UNSUPPORTED_PARAM_TYPE 2035

-#define CR_SHARED_MEMORY_CONNECTION             2035
-#define CR_SHARED_MEMORY_CONNECT_REQUEST_ERROR  2036
-#define CR_SHARED_MEMORY_CONNECT_ANSWER_ERROR   2037
-#define CR_SHARED_MEMORY_CONNECT_FILE_MAP_ERROR  2038
-#define CR_SHARED_MEMORY_CONNECT_MAP_ERROR      2039
-#define CR_SHARED_MEMORY_FILE_MAP_ERROR         2040
-#define CR_SHARED_MEMORY_MAP_ERROR              2041
-#define CR_SHARED_MEMORY_EVENT_ERROR     	2042
-#define CR_SHARED_MEMORY_CONNECT_ABANDODED_ERROR  2043
-#define CR_SHARED_MEMORY_CONNECT_SET_ERROR      2046
-#define CR_CONN_UNKNOW_PROTOCOL 		2048
+#define CR_SHARED_MEMORY_CONNECTION             2036
+#define CR_SHARED_MEMORY_CONNECT_REQUEST_ERROR  2037
+#define CR_SHARED_MEMORY_CONNECT_ANSWER_ERROR   2038
+#define CR_SHARED_MEMORY_CONNECT_FILE_MAP_ERROR 2039
+#define CR_SHARED_MEMORY_CONNECT_MAP_ERROR      2040
+#define CR_SHARED_MEMORY_FILE_MAP_ERROR         2041
+#define CR_SHARED_MEMORY_MAP_ERROR              2042
+#define CR_SHARED_MEMORY_EVENT_ERROR     	2043
+#define CR_SHARED_MEMORY_CONNECT_ABANDODED_ERROR 2044
+#define CR_SHARED_MEMORY_CONNECT_SET_ERROR      2045
+#define CR_CONN_UNKNOW_PROTOCOL 		2046
--- a/include/my_alloc.h
+++ b/include/my_alloc.h
@ -21,7 +21,8 @@
 #ifndef _my_alloc_h
 #define _my_alloc_h

-#define MAX_BLOCK_USAGE_BEFORE_DROP 10
+#define ALLOC_MAX_BLOCK_TO_DROP			4096
+#define ALLOC_MAX_BLOCK_USAGE_BEFORE_DROP	10

 typedef struct st_used_mem
 {				   /* struct for once_alloc (block) */
--- a/include/my_base.h
+++ b/include/my_base.h
@ -301,8 +301,10 @@ enum data_file_type {

 /* For number of records */
 #ifdef BIG_TABLES
+#define rows2double(A)	ulonglong2double(A)
 typedef my_off_t	ha_rows;
 #else
+#define rows2double(A)	(double) (A)
 typedef ulong		ha_rows;
 #endif

--- a/include/my_global.h
+++ b/include/my_global.h
@ -38,6 +38,14 @@
 #define HAVE_ERRNO_AS_DEFINE
 #endif /* __CYGWIN__ */

+/* Macros to make switching between C and C++ mode easier */
+#ifdef __cplusplus
+#define C_MODE_START    extern "C" {
+#define C_MODE_END	}
+#else
+#define C_MODE_START
+#define C_MODE_END
+#endif

 #if defined(_WIN32) || defined(_WIN64) || defined(__WIN32__) || defined(WIN32)
 #include <config-win.h>
@ -119,7 +127,12 @@
 #define _H_STRINGS
 #define _SYS_STREAM_H
 /* #define _AIX32_CURSES */	/* XXX: this breaks AIX 4.3.3 (others?). */
-#endif
+#define ulonglong2double(A) my_ulonglong2double(A)
+#define my_off_t2double(A)  my_ulonglong2double(A)
+C_MODE_START
+double my_ulonglong2double(unsigned long long A);
+C_MODE_END
+#endif /* _AIX */

 #ifdef HAVE_BROKEN_SNPRINTF	/* HPUX 10.20 don't have this defined */
 #undef HAVE_SNPRINTF
@ -128,6 +141,10 @@
 #undef HAVE_PREAD
 #undef HAVE_PWRITE
 #endif
+#if defined(HAVE_BROKEN_INLINE) && !defined(__cplusplus)
+#undef inline
+#define inline
+#endif

 #ifdef UNDEF_HAVE_GETHOSTBYNAME_R		/* For OSF4.x */
 #undef HAVE_GETHOSTBYNAME_R
@ -240,7 +257,7 @@
 #ifdef DONT_USE_FINITE		/* HPUX 11.x has is_finite() */
 #undef HAVE_FINITE
 #endif
-#if defined(HPUX) && defined(_LARGEFILE64_SOURCE) && defined(THREAD)
+#if defined(HPUX10) && defined(_LARGEFILE64_SOURCE) && defined(THREAD)
 /* Fix bug in setrlimit */
 #undef setrlimit
 #define setrlimit cma_setrlimit64
@ -376,7 +393,9 @@ typedef int	my_socket;	/* File descriptor for sockets */
 #endif
 /* Type for fuctions that handles signals */
 #define sig_handler RETSIGTYPE
+C_MODE_START
 typedef void	(*sig_return)();/* Returns type from signal */
+C_MODE_END
 #if defined(__GNUC__) && !defined(_lint)
 typedef char	pchar;		/* Mixed prototypes can take char */
 typedef char	puchar;		/* Mixed prototypes can take char */
@ -390,8 +409,10 @@ typedef int	pbool;		/* Mixed prototypes can't take char */
 typedef int	pshort;		/* Mixed prototypes can't take short int */
 typedef double	pfloat;		/* Mixed prototypes can't take float */
 #endif
+C_MODE_START
 typedef int	(*qsort_cmp)(const void *,const void *);
 typedef int	(*qsort_cmp2)(void*, const void *,const void *);
+C_MODE_END
 #ifdef HAVE_mit_thread
 #define qsort_t void
 #undef QSORT_TYPE_IS_VOID
@ -1029,13 +1050,4 @@ typedef union {
 #define statistic_add(V,C,L)       (V)+=(C)
 #endif

-/* Macros to make switching between C and C++ mode easier */
-#ifdef __cplusplus
-#define C_MODE_START    extern "C" {
-#define C_MODE_END	}
-#else
-#define C_MODE_START
-#define C_MODE_END
-#endif
-
-#endif /* _global_h */
+#endif /* my_global_h */
--- a/include/my_net.h
+++ b/include/my_net.h
@ -71,7 +71,7 @@ void my_inet_ntoa(struct in_addr in, char *buf);
  Handling of gethostbyname_r()
 */

-#if !defined(HPUX)
+#if !defined(HPUX10)
 struct hostent;
 #endif /* HPUX */
 #if !defined(HAVE_GETHOSTBYNAME_R)
@ -84,7 +84,7 @@ struct hostent *my_gethostbyname_r(const char *name,
 				   struct hostent *result, char *buffer,
 				   int buflen, int *h_errnop);
 #define my_gethostbyname_r_free()
-#if !defined(HAVE_GETHOSTBYNAME_R_GLIBC2_STYLE) && !defined(HPUX)
+#if !defined(HAVE_GETHOSTBYNAME_R_GLIBC2_STYLE) && !defined(HPUX10)
 #define GETHOSTBYNAME_BUFF_SIZE sizeof(struct hostent_data)
 #endif /* !defined(HAVE_GETHOSTBYNAME_R_GLIBC2_STYLE) */

--- a/include/my_pthread.h
+++ b/include/my_pthread.h
@ -428,7 +428,7 @@ struct tm *localtime_r(const time_t *clock, struct tm *res);

 #endif /* defined(__WIN__) */

-#if defined(HPUX) && !defined(DONT_REMAP_PTHREAD_FUNCTIONS)
+#if defined(HPUX10) && !defined(DONT_REMAP_PTHREAD_FUNCTIONS)
 #undef pthread_cond_timedwait
 #define pthread_cond_timedwait(a,b,c) my_pthread_cond_timedwait((a),(b),(c))
 int my_pthread_cond_timedwait(pthread_cond_t *cond, pthread_mutex_t *mutex,
--- a/include/my_tree.h
+++ b/include/my_tree.h
@ -83,7 +83,6 @@ void *tree_search(TREE *tree, void *key, void *custom_arg);
 int tree_walk(TREE *tree,tree_walk_action action,
 	      void *argument, TREE_WALK visit);
 int tree_delete(TREE *tree, void *key, void *custom_arg);
-
 void *tree_search_key(TREE *tree, const void *key, 
                      TREE_ELEMENT **parents, TREE_ELEMENT ***last_pos,
                      enum ha_rkey_function flag, void *custom_arg);
@ -93,6 +92,9 @@ void *tree_search_next(TREE *tree, TREE_ELEMENT ***last_pos, int l_offs,
                       int r_offs);
 ha_rows tree_record_pos(TREE *tree, const void *key, 
                     enum ha_rkey_function search_flag, void *custom_arg);
+
+#define TREE_ELEMENT_EXTRA_SIZE (sizeof(TREE_ELEMENT) + sizeof(void*))
+
 #ifdef	__cplusplus
 }
 #endif
--- a/include/myisam.h
+++ b/include/myisam.h
@ -285,7 +285,7 @@ extern uint mi_get_pointer_length(ulonglong file_length, uint def);
 #define T_VERBOSE               (1L << 28)
 #define T_VERY_SILENT           (1L << 29)
 #define T_WAIT_FOREVER          (1L << 30)
-#define T_WRITE_LOOP            (1L << 31)
+#define T_WRITE_LOOP            ((ulong) 1L << 31)

 #define T_REP_ANY               (T_REP | T_REP_BY_SORT | T_REP_PARALLEL)

--- a/include/queues.h
+++ b/include/queues.h
@ -41,12 +41,13 @@ typedef struct st_queue {
 #define queue_element(queue,index) ((queue)->root[index+1])
 #define queue_end(queue) ((queue)->root[(queue)->elements])
 #define queue_replaced(queue) _downheap(queue,1)
+typedef int (*queue_compare)(void *,byte *, byte *);

 int init_queue(QUEUE *queue,uint max_elements,uint offset_to_key,
-	       pbool max_at_top, int (*compare)(void *,byte *, byte *),
+	       pbool max_at_top, queue_compare compare,
 	       void *first_cmp_arg);
 int reinit_queue(QUEUE *queue,uint max_elements,uint offset_to_key,
-                 pbool max_at_top, int (*compare)(void *,byte *, byte *),
+                 pbool max_at_top, queue_compare compare,
                 void *first_cmp_arg);
 void delete_queue(QUEUE *queue);
 void queue_insert(QUEUE *queue,byte *element);
--- a/include/violite.h
+++ b/include/violite.h
@ -188,7 +188,7 @@ struct st_VioSSLConnectorFd
  SSL_METHOD* ssl_method_;
 };

-void sslaccept(struct st_VioSSLAcceptorFd*, Vio*, long timeout);
+int sslaccept(struct st_VioSSLAcceptorFd*, Vio*, long timeout);
 int sslconnect(struct st_VioSSLConnectorFd*, Vio*, long timeout);

 struct st_VioSSLConnectorFd
@ -245,7 +245,6 @@ struct st_vio

 #ifdef HAVE_OPENSSL
  SSL* ssl_;
-  my_bool open_;
 #endif /* HAVE_OPENSSL */
 #ifdef HAVE_SMEM
  HANDLE handle_file_map;
--- a/innobase/btr/btr0btr.c
+++ b/innobase/btr/btr0btr.c
@ -274,6 +274,7 @@ btr_page_create(
 	ut_ad(mtr_memo_contains(mtr, buf_block_align(page),
 			      				MTR_MEMO_PAGE_X_FIX));
 	page_create(page, mtr);
+	buf_block_align(page)->check_index_page_at_flush = TRUE;
 	
 	btr_page_set_index_id(page, tree->id, mtr);
 }
@ -713,6 +714,7 @@ btr_create(
 	
 	/* Create a new index page on the the allocated segment page */
 	page = page_create(frame, mtr);
+	buf_block_align(page)->check_index_page_at_flush = TRUE;

 	/* Set the index id of the page */
 	btr_page_set_index_id(page, index_id, mtr);
@ -847,6 +849,7 @@ btr_page_reorganize_low(
 	segment headers, next page-field, etc.) is preserved intact */

 	page_create(page, mtr);
+	buf_block_align(page)->check_index_page_at_flush = TRUE;
 	
 	/* Copy the records from the temporary space to the recreated page;
 	do not copy the lock bits yet */
@ -919,6 +922,7 @@ btr_page_empty(
 	segment headers, next page-field, etc.) is preserved intact */

 	page_create(page, mtr);
+	buf_block_align(page)->check_index_page_at_flush = TRUE;
 }

 /*****************************************************************
@ -2291,29 +2295,54 @@ btr_check_node_ptr(
 /****************************************************************
 Checks the size and number of fields in a record based on the definition of
 the index. */
-static
+
 ibool
 btr_index_rec_validate(
 /*====================*/
 					/* out: TRUE if ok */
 	rec_t*		rec,		/* in: index record */
-	dict_index_t*	index)	/* in: index */
+	dict_index_t*	index,		/* in: index */
+	ibool		dump_on_error)	/* in: TRUE if the function
+					should print hex dump of record
+					and page on error */
 {
 	dtype_t* type;
 	byte*	data;
 	ulint	len;
 	ulint	n;
 	ulint	i;
+	page_t*	page;
 	char	err_buf[1000];

+	page = buf_frame_align(rec);
+	
+	if (index->type & DICT_UNIVERSAL) {
+	        /* The insert buffer index tree can contain records from any
+	        other index: we cannot check the number of fields or
+	        their length */
+
+	        return(TRUE);
+	}
+
 	n = dict_index_get_n_fields(index);

 	if (rec_get_n_fields(rec) != n) {
-		fprintf(stderr, "Record has %lu fields, should have %lu\n",
+		fprintf(stderr,
+"InnoDB: Record in index %s in table %s, page %lu, at offset %lu\n"
+"InnoDB: has %lu fields, should have %lu\n",
+			index->name, index->table_name,
+			buf_frame_get_page_no(page), (ulint)(rec - page),
 			rec_get_n_fields(rec), n);

+		if (!dump_on_error) {
+
+			return(FALSE);
+		}
+
+		buf_page_print(page);
+
 		rec_sprintf(err_buf, 900, rec);
-	  	fprintf(stderr, "InnoDB: record %s\n", err_buf);
+	  	fprintf(stderr, "InnoDB: corrupt record %s\n", err_buf);

 		return(FALSE);
 	}
@ -2326,11 +2355,23 @@ btr_index_rec_validate(
 		if (len != UNIV_SQL_NULL && dtype_is_fixed_size(type)
 		    			&& len != dtype_get_fixed_size(type)) {
 			fprintf(stderr,
-			"Record field %lu len is %lu, should be %lu\n",
+"InnoDB: Record in index %s in table %s, page %lu, at offset %lu\n"
+"InnoDB: field %lu len is %lu, should be %lu\n",
+				index->name, index->table_name,
+				buf_frame_get_page_no(page),
+				(ulint)(rec - page),
 				i, len, dtype_get_fixed_size(type));

+			if (!dump_on_error) {
+	
+				return(FALSE);
+			}
+
+			buf_page_print(page);
+
 			rec_sprintf(err_buf, 900, rec);
-	  		fprintf(stderr, "InnoDB: record %s\n", err_buf);
+	  		fprintf(stderr,
+                             "InnoDB: corrupt record %s\n", err_buf);

 			return(FALSE);
 		}
@ -2361,12 +2402,13 @@ btr_index_page_validate(
 		rec = (&cur)->rec;

 		if (page_cur_is_after_last(&cur)) {
+
 			break;
 		}

-		if (!btr_index_rec_validate(rec, index)) {
+		if (!btr_index_rec_validate(rec, index, TRUE)) {

-			ret = FALSE;
+			return(FALSE);
 		}

 		page_cur_move_to_next(&cur);
@ -2423,25 +2465,26 @@ btr_validate_level(

 	index = UT_LIST_GET_FIRST(tree->tree_indexes);
 	
-	/* Now we are on the desired level */
+	/* Now we are on the desired level. Loop through the pages on that
+	level. */
 loop:
 	mtr_x_lock(dict_tree_get_lock(tree), &mtr);

 	/* Check ordering etc. of records */

 	if (!page_validate(page, index)) {
-		fprintf(stderr, "Error in page %lu in index %s\n",
-			buf_frame_get_page_no(page), index->name);
+		fprintf(stderr,
+"InnoDB: Error in page %lu in index %s table %s, index tree level %lu\n",
+			buf_frame_get_page_no(page), index->name,
+			index->table_name, level);

 		ret = FALSE;
-	}
+	} else if (level == 0) {
+		/* We are on level 0. Check that the records have the right
+		number of fields, and field lengths are right. */

-	if (level == 0) {
 		if (!btr_index_page_validate(page, index)) {
- 			fprintf(stderr,
-				"Error in page %lu in index %s, level %lu\n",
-				buf_frame_get_page_no(page), index->name,
-								level);
+
 			ret = FALSE;
 		}
 	}
@ -2464,14 +2507,17 @@ loop:
 			UT_LIST_GET_FIRST(tree->tree_indexes)) >= 0) {

 			fprintf(stderr,
-			"InnoDB: Error on pages %lu and %lu in index %s\n",
+		"InnoDB: Error on pages %lu and %lu in index %s table %s\n",
 				buf_frame_get_page_no(page),
 				right_page_no,
-				index->name);
+				index->name, index->table_name);

 			fprintf(stderr,
 			"InnoDB: records in wrong order on adjacent pages\n");

+			buf_page_print(page);
+			buf_page_print(right_page);
+
 			rec_sprintf(err_buf, 900,
 				page_rec_get_prev(page_get_supremum_rec(page)));
 	  		fprintf(stderr, "InnoDB: record %s\n", err_buf);
@ -2494,6 +2540,7 @@ loop:
 		/* Check father node pointers */
 	
 		node_ptr = btr_page_get_father_node_ptr(tree, page, &mtr);
+		father_page = buf_frame_align(node_ptr);

 		if (btr_node_ptr_get_child_page_no(node_ptr) !=
 						buf_frame_get_page_no(page)
@ -2501,13 +2548,16 @@ loop:
 		   	page_rec_get_prev(page_get_supremum_rec(page)),
 								&mtr)) {
 			fprintf(stderr,
-			"InnoDB: Error on page %lu in index %s\n",
+			"InnoDB: Error on page %lu in index %s table %s\n",
 				buf_frame_get_page_no(page),
-				index->name);
+				index->name, index->table_name);

 			fprintf(stderr,
 			"InnoDB: node pointer to the page is wrong\n");

+			buf_page_print(father_page);
+			buf_page_print(page);
+
 			rec_sprintf(err_buf, 900, node_ptr);
 				
 	  		fprintf(stderr, "InnoDB: node ptr %s\n", err_buf);
@ -2528,8 +2578,6 @@ loop:
 		   	goto node_ptr_fails;
 		}

-		father_page = buf_frame_align(node_ptr);
-
 		if (btr_page_get_level(page, &mtr) > 0) {
 			heap = mem_heap_create(256);
 		
@ -2543,9 +2591,12 @@ loop:
 			if (cmp_dtuple_rec(node_ptr_tuple, node_ptr) != 0) {

 	 			fprintf(stderr,
-				  "InnoDB: Error on page %lu in index %s\n",
+			"InnoDB: Error on page %lu in index %s table %s\n",
 					buf_frame_get_page_no(page),
-					index->name);
+					index->name, index->table_name);
+
+				buf_page_print(father_page);
+				buf_page_print(page);

 	  			fprintf(stderr,
                	"InnoDB: Error: node ptrs differ on levels > 0\n");
@ -2595,9 +2646,13 @@ loop:
 			"InnoDB: node pointer to the right page is wrong\n");

 	 				fprintf(stderr,
-				  "InnoDB: Error on page %lu in index %s\n",
+			"InnoDB: Error on page %lu in index %s table %s\n",
 					buf_frame_get_page_no(page),
-					index->name);
+					index->name, index->table_name);
+
+					buf_page_print(father_page);
+					buf_page_print(page);
+					buf_page_print(right_page);
 				}
 			} else {
 				right_father_page = buf_frame_align(
@ -2611,9 +2666,14 @@ loop:
 			"InnoDB: node pointer 2 to the right page is wrong\n");

 	 				fprintf(stderr,
-				  "InnoDB: Error on page %lu in index %s\n",
+			"InnoDB: Error on page %lu in index %s table %s\n",
 					buf_frame_get_page_no(page),
-					index->name);
+					index->name, index->table_name);
+
+					buf_page_print(father_page);
+					buf_page_print(right_father_page);
+					buf_page_print(page);
+					buf_page_print(right_page);
 				}

 				if (buf_frame_get_page_no(right_father_page)
@ -2624,9 +2684,14 @@ loop:
 			"InnoDB: node pointer 3 to the right page is wrong\n");

 	 				fprintf(stderr,
-				  "InnoDB: Error on page %lu in index %s\n",
+			"InnoDB: Error on page %lu in index %s table %s\n",
 					buf_frame_get_page_no(page),
-					index->name);
+					index->name, index->table_name);
+
+					buf_page_print(father_page);
+					buf_page_print(right_father_page);
+					buf_page_print(page);
+					buf_page_print(right_page);
 				}
 			}					
 		}
--- a/innobase/btr/btr0cur.c
+++ b/innobase/btr/btr0cur.c
@ -121,16 +121,19 @@ btr_cur_latch_leaves(
 {
 	ulint	left_page_no;
 	ulint	right_page_no;
+	page_t*	get_page;
 	
 	ut_ad(tree && page && mtr);

 	if (latch_mode == BTR_SEARCH_LEAF) {
 	
-		btr_page_get(space, page_no, RW_S_LATCH, mtr);
+		get_page = btr_page_get(space, page_no, RW_S_LATCH, mtr);
+		buf_block_align(get_page)->check_index_page_at_flush = TRUE;

 	} else if (latch_mode == BTR_MODIFY_LEAF) {

-		btr_page_get(space, page_no, RW_X_LATCH, mtr);
+		get_page = btr_page_get(space, page_no, RW_X_LATCH, mtr);
+		buf_block_align(get_page)->check_index_page_at_flush = TRUE;

 	} else if (latch_mode == BTR_MODIFY_TREE) {

@ -138,15 +141,22 @@ btr_cur_latch_leaves(
 		left_page_no = btr_page_get_prev(page, mtr);

 		if (left_page_no != FIL_NULL) {
-			btr_page_get(space, left_page_no, RW_X_LATCH, mtr);
+			get_page = btr_page_get(space, left_page_no,
+							RW_X_LATCH, mtr);
+			buf_block_align(get_page)->check_index_page_at_flush =
+									TRUE;
 		}
 				
-		btr_page_get(space, page_no, RW_X_LATCH, mtr);
+		get_page = btr_page_get(space, page_no, RW_X_LATCH, mtr);
+		buf_block_align(get_page)->check_index_page_at_flush = TRUE;

 		right_page_no = btr_page_get_next(page, mtr);

 		if (right_page_no != FIL_NULL) {
-			btr_page_get(space, right_page_no, RW_X_LATCH, mtr);
+			get_page = btr_page_get(space, right_page_no,
+							RW_X_LATCH, mtr);
+			buf_block_align(get_page)->check_index_page_at_flush =
+									TRUE;
 		}

 	} else if (latch_mode == BTR_SEARCH_PREV) {
@ -157,9 +167,12 @@ btr_cur_latch_leaves(
 		if (left_page_no != FIL_NULL) {
 			cursor->left_page = btr_page_get(space, left_page_no,
 							RW_S_LATCH, mtr);
+			buf_block_align(
+			cursor->left_page)->check_index_page_at_flush = TRUE;
 		}

-		btr_page_get(space, page_no, RW_S_LATCH, mtr);
+		get_page = btr_page_get(space, page_no, RW_S_LATCH, mtr);
+		buf_block_align(get_page)->check_index_page_at_flush = TRUE;

 	} else if (latch_mode == BTR_MODIFY_PREV) {

@ -169,9 +182,12 @@ btr_cur_latch_leaves(
 		if (left_page_no != FIL_NULL) {
 			cursor->left_page = btr_page_get(space, left_page_no,
 							RW_X_LATCH, mtr);
+			buf_block_align(
+			cursor->left_page)->check_index_page_at_flush = TRUE;
 		}

-		btr_page_get(space, page_no, RW_X_LATCH, mtr);
+		get_page = btr_page_get(space, page_no, RW_X_LATCH, mtr);
+		buf_block_align(get_page)->check_index_page_at_flush = TRUE;
 	} else {
 		ut_error;
 	}
@ -274,6 +290,7 @@ btr_cur_search_to_nth_level(
 	if (btr_search_latch.writer == RW_LOCK_NOT_LOCKED
 		&& latch_mode <= BTR_MODIFY_LEAF && info->last_hash_succ
 		&& !estimate
+		&& mode != PAGE_CUR_LE_OR_EXTENDS
 	        && btr_search_guess_on_hash(index, info, tuple, mode,
 						latch_mode, cursor,
 						has_search_latch, mtr)) {
@ -334,12 +351,18 @@ btr_cur_search_to_nth_level(
 	rw_latch = RW_NO_LATCH;
 	buf_mode = BUF_GET;

+	/* We use these modified search modes on non-leaf levels of the
+	B-tree. These let us end up in the right B-tree leaf. In that leaf
+	we use the original search mode. */
+
 	if (mode == PAGE_CUR_GE) {
 		page_mode = PAGE_CUR_L;
 	} else if (mode == PAGE_CUR_G) {
 		page_mode = PAGE_CUR_LE;
 	} else if (mode == PAGE_CUR_LE) {
 		page_mode = PAGE_CUR_LE;
+	} else if (mode == PAGE_CUR_LE_OR_EXTENDS) {
+		page_mode = PAGE_CUR_LE_OR_EXTENDS;
 	} else {
 		ut_ad(mode == PAGE_CUR_L);
 		page_mode = PAGE_CUR_L;
@ -391,6 +414,8 @@ retry_page_get:
 			goto retry_page_get;
 		}

+		buf_block_align(page)->check_index_page_at_flush = TRUE;
+			
 #ifdef UNIV_SYNC_DEBUG					
 		if (rw_latch != RW_NO_LATCH) {
 			buf_page_dbg_add_level(page, SYNC_TREE_NODE);
@ -543,6 +568,8 @@ btr_cur_open_at_index_side(
 		ut_ad(0 == ut_dulint_cmp(tree->id,
 						btr_page_get_index_id(page)));

+		buf_block_align(page)->check_index_page_at_flush = TRUE;
+
 		if (height == ULINT_UNDEFINED) {
 			/* We are in the root node */

--- a/innobase/btr/btr0pcur.c
+++ b/innobase/btr/btr0pcur.c
@ -354,6 +354,7 @@ btr_pcur_move_to_next_page(
 	ut_ad(next_page_no != FIL_NULL);	

 	next_page = btr_page_get(space, next_page_no, cursor->latch_mode, mtr);
+	buf_block_align(next_page)->check_index_page_at_flush = TRUE;

 	btr_leaf_page_release(page, cursor->latch_mode, mtr);
 	
--- a/innobase/buf/buf0buf.c
+++ b/innobase/buf/buf0buf.c
@ -331,6 +331,11 @@ buf_page_print(
 						index->table_name,
 						index->name);
 		}
+	} else if (fil_page_get_type(read_buf) == FIL_PAGE_INODE) {
+		fprintf(stderr, "InnoDB: Page may be an 'inode' page\n");
+	} else if (fil_page_get_type(read_buf) == FIL_PAGE_IBUF_FREE_LIST) {
+		fprintf(stderr,
+		"InnoDB: Page may be an insert buffer free list page\n");
 	}
 }

@ -351,6 +356,8 @@ buf_block_init(
 	
 	block->file_page_was_freed = FALSE;

+	block->check_index_page_at_flush = FALSE;
+
 	rw_lock_create(&(block->lock));
 	ut_ad(rw_lock_validate(&(block->lock)));

@ -616,6 +623,29 @@ buf_page_peek_block(
 	return(block);
 }

+/************************************************************************
+Resets the check_index_page_at_flush field of a page if found in the buffer
+pool. */
+
+void
+buf_reset_check_index_page_at_flush(
+/*================================*/
+	ulint	space,	/* in: space id */
+	ulint	offset)	/* in: page number */
+{
+	buf_block_t*	block;
+
+	mutex_enter_fast(&(buf_pool->mutex));
+
+	block = buf_page_hash_get(space, offset);
+
+	if (block) {
+		block->check_index_page_at_flush = FALSE;
+	}
+	
+	mutex_exit(&(buf_pool->mutex));
+}
+
 /************************************************************************
 Returns the current state of is_hashed of a page. FALSE if the page is
 not in the pool. NOTE that this operation does not fix the page in the
@ -1185,6 +1215,8 @@ buf_page_init(
 	block->space 		= space;
 	block->offset 		= offset;

+	block->check_index_page_at_flush = FALSE;
+	
 	block->lock_hash_val	= lock_rec_hash(space, offset);
 	block->lock_mutex	= NULL;
 	
@ -1326,11 +1358,6 @@ buf_page_create(

 	free_block = buf_LRU_get_free_block();
 	
-	/* Delete possible entries for the page from the insert buffer:
-	such can exist if the page belonged to an index which was dropped */
-
-	ibuf_merge_or_delete_for_page(NULL, space, offset);	
-	
 	mutex_enter(&(buf_pool->mutex));

 	block = buf_page_hash_get(space, offset);
@ -1378,6 +1405,11 @@ buf_page_create(

 	mutex_exit(&(buf_pool->mutex));

+	/* Delete possible entries for the page from the insert buffer:
+	such can exist if the page belonged to an index which was dropped */
+
+	ibuf_merge_or_delete_for_page(NULL, space, offset);	
+
 	/* Flush pages from the end of the LRU list if necessary */
 	buf_flush_free_margin();

--- a/innobase/buf/buf0flu.c
+++ b/innobase/buf/buf0flu.c
@ -15,6 +15,7 @@ Created 11/11/1995 Heikki Tuuri

 #include "ut0byte.h"
 #include "ut0lst.h"
+#include "page0page.h"
 #include "fil0fil.h"
 #include "buf0buf.h"
 #include "buf0lru.h"
@ -225,6 +226,24 @@ buf_flush_buffered_writes(void)
 		return;
 	}

+	for (i = 0; i < trx_doublewrite->first_free; i++) {
+		block = trx_doublewrite->buf_block_arr[i];
+
+		if (block->check_index_page_at_flush
+				&& !page_simple_validate(block->frame)) {
+
+			buf_page_print(block->frame);
+
+			ut_print_timestamp(stderr);
+			fprintf(stderr,
+	"  InnoDB: Apparent corruption of an index page\n"
+	"InnoDB: to be written to data file. We intentionally crash server\n"
+	"InnoDB: to prevent corrupt data from ending up in data\n"
+	"InnoDB: files.\n");
+			ut_a(0);
+		}
+	}
+
 	if (trx_doublewrite->first_free > TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
 		len = TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * UNIV_PAGE_SIZE;
 	} else {
--- a/innobase/configure.in
+++ b/innobase/configure.in
@ -86,6 +86,8 @@ else
 fi

 case "$target_os" in
+       lin*)
+	 CFLAGS="$CFLAGS -DUNIV_LINUX";;
       hpux10*)
 	 CFLAGS="$CFLAGS -DUNIV_MUST_NOT_INLINE -DUNIV_HPUX -DUNIV_HPUX10";;
       hp*)
--- a/innobase/dict/dict0crea.c
+++ b/innobase/dict/dict0crea.c
@ -1060,20 +1060,24 @@ dict_create_or_check_foreign_constraint_tables(void)
            	return(DB_SUCCESS);
        }

+	mutex_exit(&(dict_sys->mutex));
+
 	trx = trx_allocate_for_mysql();
 	
 	trx->op_info = (char *) "creating foreign key sys tables";

+	row_mysql_lock_data_dictionary(trx);
+
 	if (table1) {
 		fprintf(stderr,
 		"InnoDB: dropping incompletely created SYS_FOREIGN table\n");
-		row_drop_table_for_mysql((char *) "SYS_FOREIGN", trx, TRUE);
+		row_drop_table_for_mysql((char *) "SYS_FOREIGN", trx);
 	}

 	if (table2) {
 		fprintf(stderr,
 	"InnoDB: dropping incompletely created SYS_FOREIGN_COLS table\n");
-		row_drop_table_for_mysql((char *) "SYS_FOREIGN_COLS",trx,TRUE);
+		row_drop_table_for_mysql((char *) "SYS_FOREIGN_COLS", trx);
 	}

 	fprintf(stderr,
@ -1122,8 +1126,8 @@ dict_create_or_check_foreign_constraint_tables(void)
 		fprintf(stderr,
 		"InnoDB: dropping incompletely created SYS_FOREIGN tables\n");

-		row_drop_table_for_mysql((char *) "SYS_FOREIGN", trx, TRUE);
-		row_drop_table_for_mysql((char *) "SYS_FOREIGN_COLS",trx,TRUE);
+		row_drop_table_for_mysql((char *) "SYS_FOREIGN", trx);
+		row_drop_table_for_mysql((char *) "SYS_FOREIGN_COLS", trx);

 		error = DB_MUST_GET_MORE_FILE_SPACE;
 	}
@ -1132,6 +1136,8 @@ dict_create_or_check_foreign_constraint_tables(void)
 	
 	trx->op_info = (char *) "";

+	row_mysql_unlock_data_dictionary(trx);
+
  	trx_free_for_mysql(trx);

  	if (error == DB_SUCCESS) {
@ -1139,8 +1145,6 @@ dict_create_or_check_foreign_constraint_tables(void)
 		"InnoDB: Foreign key constraint system tables created\n");
 	}

-	mutex_exit(&(dict_sys->mutex));
-
 	return(error);
 }

--- a/innobase/dict/dict0dict.c
+++ b/innobase/dict/dict0dict.c
@ -29,7 +29,17 @@ Created 1/8/1996 Heikki Tuuri

 dict_sys_t*	dict_sys	= NULL;	/* the dictionary system */

-rw_lock_t	dict_foreign_key_check_lock;
+rw_lock_t	dict_operation_lock;	/* table create, drop, etc. reserve
+					this in X-mode; implicit or backround
+					operations purge, rollback, foreign
+					key checks reserve this in S-mode; we
+					cannot trust that MySQL protects
+					implicit or background operations
+					a table drop since MySQL does not
+					know of them; therefore we need this;
+					NOTE: a transaction which reserves
+					this must keep book on the mode in
+					trx->dict_operation_lock_mode */

 #define	DICT_HEAP_SIZE		100	/* initial memory heap size when
 					creating a table or index object */
@ -175,6 +185,58 @@ dict_foreign_free(
 /*==============*/
 	dict_foreign_t*	foreign);	/* in, own: foreign key struct */

+/************************************************************************
+Checks if the database name in two table names is the same. */
+static
+ibool
+dict_tables_have_same_db(
+/*=====================*/
+			/* out: TRUE if same db name */
+	char*	name1,	/* in: table name in the form dbname '/' tablename */
+	char*	name2)	/* in: table name in the form dbname '/' tablename */
+{
+	ulint	i;
+
+	for (i = 0; i < 100000; i++) {
+		if (name1[i] == '/' && name2[i] == '/') {
+
+			return(TRUE);
+		}
+
+		if (name1[i] != name2[i]) {
+
+			return(FALSE);
+		}
+	}
+
+	ut_a(0);
+
+	return(FALSE);
+}
+
+/************************************************************************
+Return the end of table name where we have removed dbname and '/'. */
+static
+char*
+dict_remove_db_name(
+/*================*/
+			/* out: table name */
+	char*	name)	/* in: table name in the form dbname '/' tablename */
+{
+	ulint	i;
+
+	for (i = 0; i < 100000 ; i++) {
+		if (name[i] == '/') {
+
+			return(name + i + 1);
+		}
+	}
+
+	ut_a(0);
+
+	return(NULL);
+}
+	
 /************************************************************************
 Reserves the dictionary system mutex for MySQL. */

@ -509,9 +571,8 @@ dict_init(void)

 	UT_LIST_INIT(dict_sys->table_LRU);

-	rw_lock_create(&dict_foreign_key_check_lock);
-	rw_lock_set_level(&dict_foreign_key_check_lock,
-						SYNC_FOREIGN_KEY_CHECK);
+	rw_lock_create(&dict_operation_lock);
+	rw_lock_set_level(&dict_operation_lock, SYNC_DICT_OPERATION);
 }

 /**************************************************************************
@ -1851,14 +1912,14 @@ loop:

 /*************************************************************************
 Accepts a specified string. Comparisons are case-insensitive. */
-static
+
 char*
 dict_accept(
 /*========*/
 			/* out: if string was accepted, the pointer
 			is moved after that, else ptr is returned */
 	char*	ptr,	/* in: scan from this */
-	const char* string,	/* in: accept only this string as the next
+	const char* string,/* in: accept only this string as the next
 			non-whitespace string */
 	ibool*	success)/* out: TRUE if accepted */
 {
@ -1920,7 +1981,8 @@ dict_scan_col(

 	old_ptr = ptr;
 	
-	while (!isspace(*ptr) && *ptr != ',' && *ptr != ')' && 	*ptr != '`') {
+	while (!isspace(*ptr) && *ptr != ',' && *ptr != ')' && 	*ptr != '`'
+	       && *ptr != '\0') {

 		ptr++;
 	}
@ -1994,7 +2056,7 @@ dict_scan_table_name(

 	old_ptr = ptr;
 	
-	while (!isspace(*ptr) && *ptr != '(' && *ptr != '`') {
+	while (!isspace(*ptr) && *ptr != '(' && *ptr != '`' && *ptr != '\0') {
 		if (*ptr == '.') {
 			dot_ptr = ptr;
 		}
@ -2019,7 +2081,13 @@ dict_scan_table_name(
 		ut_cpy_in_lower_case(second_table_name + i, old_ptr,
 				     				ptr - old_ptr);
 #else
-		ut_memcpy(second_table_name + i, old_ptr, ptr - old_ptr);
+		if (srv_lower_case_table_names) {
+			ut_cpy_in_lower_case(second_table_name + i, old_ptr,
+				     				ptr - old_ptr);
+		} else {
+			ut_memcpy(second_table_name + i, old_ptr,
+								ptr - old_ptr);
+		}
 #endif
 		second_table_name[i + (ptr - old_ptr)] = '\0';
 	} else {
@ -2027,7 +2095,12 @@ dict_scan_table_name(
 		ut_cpy_in_lower_case(second_table_name, old_ptr,
 								ptr - old_ptr);
 #else
+		if (srv_lower_case_table_names) {
+			ut_cpy_in_lower_case(second_table_name, old_ptr,
+				   			ptr - old_ptr);
+		} else {
 			ut_memcpy(second_table_name, old_ptr, ptr - old_ptr);
+		}
 #endif
 		second_table_name[dot_ptr - old_ptr] = '/';
 		second_table_name[ptr - old_ptr] = '\0';
@ -2044,6 +2117,44 @@ dict_scan_table_name(
 	return(ptr);
 }

+/*************************************************************************
+Skips one 'word', like an id. For the lexical definition of 'word', see the
+code below. */
+static
+char*
+dict_skip_word(
+/*===========*/
+			/* out: scanned to */
+	char*	ptr,	/* in: scanned to */
+	ibool*	success)/* out: TRUE if success, FALSE if just spaces left in
+			string */
+{
+	*success = FALSE;
+
+	while (isspace(*ptr)) {
+		ptr++;
+	}
+
+	if (*ptr == '\0') {
+
+		return(ptr);
+	}
+
+	if (*ptr == '`') {
+		ptr++;
+	}
+	
+	while (!isspace(*ptr) && *ptr != ',' && *ptr != '(' && 	*ptr != '`'
+	       && *ptr != '\0') {
+
+		ptr++;
+	}
+
+	*success = TRUE;
+
+	return(ptr);
+}
+
 /*************************************************************************
 Returns the number of opening brackets '(' subtracted by the number
 of closing brackets ')' between string and ptr. */
@ -2113,7 +2224,6 @@ dict_create_foreign_constraints(
 	if (table == NULL) {
 		return(DB_ERROR);
 	}
-
 loop:
 	ptr = dict_scan_to(ptr, (char *) "FOREIGN");

@ -2142,7 +2252,19 @@ loop:
 	ptr = dict_accept(ptr, (char *) "(", &success);

 	if (!success) {
-		goto loop;
+		/* MySQL allows also an index id before the '('; we
+		skip it */
+		ptr = dict_skip_word(ptr, &success);
+
+		if (!success) {
+			return(DB_CANNOT_ADD_CONSTRAINT);
+		}
+
+		ptr = dict_accept(ptr, (char *) "(", &success);
+
+		if (!success) {
+			return(DB_CANNOT_ADD_CONSTRAINT);
+		}
 	}

 	i = 0;
@ -2217,6 +2339,7 @@ col_loop1:

 	if (!success) {
 		dict_foreign_free(foreign);
+
 		return(DB_CANNOT_ADD_CONSTRAINT);
 	}

@ -2230,6 +2353,7 @@ col_loop2:
 	
 	if (!success) {
 		dict_foreign_free(foreign);
+
 		return(DB_CANNOT_ADD_CONSTRAINT);
 	}

@ -2257,14 +2381,20 @@ col_loop2:
 	ptr = dict_accept(ptr, "DELETE", &success);

 	if (!success) {
+		dict_foreign_free(foreign);
 		
+		return(DB_CANNOT_ADD_CONSTRAINT);
+	}
+
+	ptr = dict_accept(ptr, "RESTRICT", &success);
+
+	if (success) {
 		goto try_find_index;
 	}

 	ptr = dict_accept(ptr, "CASCADE", &success);

 	if (success) {
-
 		foreign->type = DICT_FOREIGN_ON_DELETE_CASCADE;

 		goto try_find_index;
@ -2273,13 +2403,19 @@ col_loop2:
 	ptr = dict_accept(ptr, "SET", &success);

 	if (!success) {
+		dict_foreign_free(foreign);
 		
-		goto try_find_index;
+		return(DB_CANNOT_ADD_CONSTRAINT);
 	}

 	ptr = dict_accept(ptr, "NULL", &success);

-	if (success) {
+	if (!success) {
+		dict_foreign_free(foreign);
+		
+		return(DB_CANNOT_ADD_CONSTRAINT);
+	}
+
 	for (j = 0; j < foreign->n_fields; j++) {
 		if ((dict_index_get_nth_type(
 				foreign->foreign_index, j)->prtype)
@ -2289,16 +2425,25 @@ col_loop2:
 			if the column is not allowed to be NULL! */

 			dict_foreign_free(foreign);
+
 			return(DB_CANNOT_ADD_CONSTRAINT);
 		}
 	}

 	foreign->type = DICT_FOREIGN_ON_DELETE_SET_NULL;
 	
-		goto try_find_index;
+try_find_index:
+	/* We check that there are no superfluous words like 'ON UPDATE ...'
+	which we do not support yet. */
+
+	ptr = dict_accept(ptr, (char *) "ON", &success);
+
+	if (success) {
+		dict_foreign_free(foreign);
+
+		return(DB_CANNOT_ADD_CONSTRAINT);
 	}
 	
-try_find_index:
 	/* Try to find an index which contains the columns as the first fields
 	and in the right order, and the types are the same as in
 	foreign->foreign_index */
@ -2345,6 +2490,7 @@ try_find_index:
 					referenced_table->referenced_list,
 								foreign);
 	}
+
 	goto loop;
 }

@ -2843,6 +2989,14 @@ dict_update_statistics_low(
 	ulint		size;
 	ulint		sum_of_index_sizes	= 0;

+	/* If we have set a high innodb_force_recovery level, do not calculate
+	statistics, as a badly corrupted index can cause a crash in it. */
+
+	if (srv_force_recovery >= SRV_FORCE_NO_IBUF_MERGE) {
+
+		return;
+	}
+
 	/* Find out the sizes of the indexes and how many different values
 	for the key they approximately have */

@ -3146,6 +3300,14 @@ dict_print_info_on_foreign_keys_in_create_format(
 			}
 		}

+		if (dict_tables_have_same_db(table->name,
+					foreign->referenced_table_name)) {
+			/* Do not print the database name of the referenced
+			table */
+			buf2 += sprintf(buf2, ") REFERENCES `%s` (",
+					dict_remove_db_name(
+					foreign->referenced_table_name));
+		} else {
 			buf2 += sprintf(buf2, ") REFERENCES `%s` (",
 					foreign->referenced_table_name);
 			/* Change the '/' in the table name to '.' */
@ -3158,6 +3320,7 @@ dict_print_info_on_foreign_keys_in_create_format(
 					break;
 				}
 			}
+		}
 	
 		for (i = 0; i < foreign->n_fields; i++) {
 		        if ((ulint)(buf2 - buf) >= len) {
--- a/innobase/dict/dict0load.c
+++ b/innobase/dict/dict0load.c
@ -335,7 +335,7 @@ dict_load_fields(
 		ut_a(btr_pcur_is_on_user_rec(&pcur, &mtr));
 		if (rec_get_deleted_flag(rec)) {
 			fprintf(stderr,
-"InnoDB: Error: data dictionary entry for table %s is corrupt!\n",
+"InnoDB: Error: data dictionary entry for table %s is corrupt!\n"
 "InnoDB: An index field is delete marked.\n",
 			table->name);
 		}
--- a/innobase/fil/fil0fil.c
+++ b/innobase/fil/fil0fil.c
@ -967,6 +967,7 @@ fil_extend_last_data_file(
 	fil_node_t*	node;
 	fil_space_t*	space;
 	fil_system_t*	system		= fil_system;
+	byte*		buf2;
 	byte*		buf;
 	ibool		success;
 	ulint		i;
@ -981,19 +982,23 @@ fil_extend_last_data_file(

 	fil_node_prepare_for_io(node, system, space);

-	buf = mem_alloc(1024 * 1024);
+	buf2 = mem_alloc(1024 * 1024 + UNIV_PAGE_SIZE);
+	buf = ut_align(buf2, UNIV_PAGE_SIZE);

 	memset(buf, '\0', 1024 * 1024);

 	for (i = 0; i < size_increase / ((1024 * 1024) / UNIV_PAGE_SIZE); i++) {

-		success = os_file_write(node->name, node->handle, buf,
+		/* If we use native Windows aio, then also this write is
+		done using it */
+
+		success = os_aio(OS_FILE_WRITE, OS_AIO_SYNC,
+			node->name, node->handle, buf,
 			(node->size << UNIV_PAGE_SIZE_SHIFT) & 0xFFFFFFFF,
 			node->size >> (32 - UNIV_PAGE_SIZE_SHIFT),
-			1024 * 1024);
+			1024 * 1024, NULL, NULL);

 		if (!success) {
-
 			break;
 		}

@ -1003,7 +1008,7 @@ fil_extend_last_data_file(
 		os_has_said_disk_full = FALSE;
 	}

-	mem_free(buf);
+	mem_free(buf2);

 	fil_node_complete_io(node, system, OS_FILE_WRITE);

@ -1528,7 +1533,6 @@ fil_page_set_type(
 	ulint	type)	/* in: type */
 {
 	ut_ad(page);
-	ut_ad((type == FIL_PAGE_INDEX) || (type == FIL_PAGE_UNDO_LOG));

 	mach_write_to_2(page + FIL_PAGE_TYPE, type);
 }	
--- a/innobase/fsp/fsp0fsp.c
+++ b/innobase/fsp/fsp0fsp.c
@ -769,6 +769,8 @@ fsp_init_file_page_low(
 #endif
 	page = buf_frame_align(ptr);

+	buf_block_align(page)->check_index_page_at_flush = FALSE;	
+	
 #ifdef UNIV_BASIC_LOG_DEBUG	
 /*	printf("In log debug version: Erase the contents of the file page\n");
 */
@ -1097,7 +1099,7 @@ fsp_fill_free_list(

 			/* Initialize the ibuf page in a separate
 			mini-transaction because it is low in the latching
-			order, and we must be able to release the its latch
+			order, and we must be able to release its latch
 			before returning from the fsp routine */
 			
 			mtr_start(&ibuf_mtr);
@ -1264,7 +1266,12 @@ fsp_alloc_free_page(

 	free = xdes_find_bit(descr, XDES_FREE_BIT, TRUE,
 						hint % FSP_EXTENT_SIZE, mtr);
-	ut_a(free != ULINT_UNDEFINED);
+	if (free == ULINT_UNDEFINED) {
+
+		ut_print_buf(((byte*)descr) - 500, 1000);
+
+		ut_a(0);
+	}

 	xdes_set_bit(descr, XDES_FREE_BIT, free, FALSE, mtr);

@ -1412,7 +1419,12 @@ fsp_free_extent(

 	descr = xdes_get_descriptor_with_space_hdr(header, space, page, mtr);

-	ut_a(xdes_get_state(descr, mtr) != XDES_FREE);
+	if (xdes_get_state(descr, mtr) == XDES_FREE) {
+
+		ut_print_buf(((byte*)descr) - 500, 1000);
+
+		ut_a(0);
+	}

 	xdes_init(descr, mtr);

@ -1523,6 +1535,10 @@ fsp_alloc_seg_inode_page(

 	page = buf_page_get(space, page_no, RW_X_LATCH, mtr);	

+	buf_block_align(page)->check_index_page_at_flush = FALSE;
+
+	fil_page_set_type(page, FIL_PAGE_INODE);
+	
 	buf_page_dbg_add_level(page, SYNC_FSP_PAGE);

 	for (i = 0; i < FSP_SEG_INODES_PER_PAGE; i++) {
@ -2298,6 +2314,8 @@ fseg_alloc_free_page_low(
 		fseg_mark_page_used(seg_inode, space, ret_page, mtr);
 	}

+	buf_reset_check_index_page_at_flush(space, ret_page);
+	
 	return(ret_page);	
 }

--- a/innobase/ibuf/ibuf0ibuf.c
+++ b/innobase/ibuf/ibuf0ibuf.c
@ -1295,6 +1295,8 @@ ibuf_add_free_page(
 	flst_add_last(root + PAGE_HEADER + PAGE_BTR_IBUF_FREE_LIST,
 		      page + PAGE_HEADER + PAGE_BTR_IBUF_FREE_LIST_NODE, &mtr);

+	fil_page_set_type(page, FIL_PAGE_IBUF_FREE_LIST);
+		      
 	ibuf_data->seg_size++;
 	ibuf_data->free_list_len++;

@ -1305,6 +1307,7 @@ ibuf_add_free_page(

 	ibuf_bitmap_page_set_bits(bitmap_page, page_no, IBUF_BITMAP_IBUF,
 								TRUE, &mtr);
+
 	mtr_commit(&mtr);

 	mutex_exit(&ibuf_mutex);
--- a/innobase/include/btr0btr.h
+++ b/innobase/include/btr0btr.h
@ -399,6 +399,19 @@ btr_print_tree(
 	dict_tree_t*	tree,	/* in: tree */
 	ulint		width);	/* in: print this many entries from start
 				and end */
+/****************************************************************
+Checks the size and number of fields in a record based on the definition of
+the index. */
+
+ibool
+btr_index_rec_validate(
+/*====================*/
+					/* out: TRUE if ok */
+	rec_t*		rec,		/* in: index record */
+	dict_index_t*	index,		/* in: index */
+	ibool		dump_on_error);	/* in: TRUE if the function
+					should print hex dump of record
+					and page on error */
 /******************************************************************
 Checks the consistency of an index tree. */

--- a/innobase/include/btr0pcur.h
+++ b/innobase/include/btr0pcur.h
@ -298,6 +298,14 @@ btr_pcur_move_to_prev(
 				function may release the page latch */
 	mtr_t*		mtr);	/* in: mtr */
 /*************************************************************
+Moves the persistent cursor to the last record on the same page. */
+UNIV_INLINE
+void
+btr_pcur_move_to_last_on_page(
+/*==========================*/
+	btr_pcur_t*	cursor,	/* in: persistent cursor */
+	mtr_t*		mtr);	/* in: mtr */
+/*************************************************************
 Moves the persistent cursor to the next user record in the tree. If no user
 records are left, the cursor ends up 'after last in tree'. */
 UNIV_INLINE
--- a/innobase/include/btr0pcur.ic
+++ b/innobase/include/btr0pcur.ic
@ -284,6 +284,24 @@ btr_pcur_move_to_prev_on_page(
 	cursor->old_stored = BTR_PCUR_OLD_NOT_STORED;
 }

+/*************************************************************
+Moves the persistent cursor to the last record on the same page. */
+UNIV_INLINE
+void
+btr_pcur_move_to_last_on_page(
+/*==========================*/
+	btr_pcur_t*	cursor,	/* in: persistent cursor */
+	mtr_t*		mtr)	/* in: mtr */
+{
+	UT_NOT_USED(mtr);
+	ut_ad(cursor->latch_mode != BTR_NO_LATCHES);
+	
+	page_cur_set_after_last(buf_frame_align(btr_pcur_get_rec(cursor)),
+				btr_pcur_get_page_cur(cursor));
+
+	cursor->old_stored = BTR_PCUR_OLD_NOT_STORED;
+}
+
 /*************************************************************
 Moves the persistent cursor to the next user record in the tree. If no user
 records are left, the cursor ends up 'after last in tree'. */
--- a/innobase/include/buf0buf.h
+++ b/innobase/include/buf0buf.h
@ -274,6 +274,15 @@ buf_page_peek_block(
 	ulint	space,	/* in: space id */
 	ulint	offset);/* in: page number */
 /************************************************************************
+Resets the check_index_page_at_flush field of a page if found in the buffer
+pool. */
+
+void
+buf_reset_check_index_page_at_flush(
+/*================================*/
+	ulint	space,	/* in: space id */
+	ulint	offset);/* in: page number */
+/************************************************************************
 Sets file_page_was_freed TRUE if the page is found in the buffer pool.
 This function should be called when we free a file page and want the
 debug version to check that it is not accessed any more unless
@ -648,6 +657,14 @@ struct buf_block_struct{
 					then it can wait for this rw-lock */
 	buf_block_t*	hash;		/* node used in chaining to the page
 					hash table */
+	ibool		check_index_page_at_flush;
+					/* TRUE if we know that this is
+					an index page, and want the database
+					to check its consistency before flush;
+					note that there may be pages in the
+					buffer pool which are index pages,
+					but this flag is not set because
+					we do not keep track of all pages */
 	/* 2. Page flushing fields */

 	UT_LIST_NODE_T(buf_block_t) flush_list;
--- a/innobase/include/db0err.h
+++ b/innobase/include/db0err.h
@ -41,6 +41,7 @@ Created 5/24/1996 Heikki Tuuri
 					which is referenced */
 #define DB_CANNOT_ADD_CONSTRAINT 38	/* adding a foreign key constraint
 					to a table failed */
+#define DB_CORRUPTION		39	/* data structure corruption noticed */
 					
 /* The following are partial failure codes */
 #define DB_FAIL 		1000
--- a/innobase/include/dict0dict.h
+++ b/innobase/include/dict0dict.h
@ -26,6 +26,18 @@ Created 1/8/1996 Heikki Tuuri
 #include "ut0byte.h"
 #include "trx0types.h"

+/*************************************************************************
+Accepts a specified string. Comparisons are case-insensitive. */
+
+char*
+dict_accept(
+/*========*/
+			/* out: if string was accepted, the pointer
+			is moved after that, else ptr is returned */
+	char*	ptr,	/* in: scan from this */
+	const char* string,/* in: accept only this string as the next
+			non-whitespace string */
+	ibool*	success);/* out: TRUE if accepted */
 /************************************************************************
 Decrements the count of open MySQL handles to a table. */

@ -798,7 +810,7 @@ dict_mutex_exit_for_mysql(void);


 extern dict_sys_t*	dict_sys;	/* the dictionary system */
-extern rw_lock_t	dict_foreign_key_check_lock;
+extern rw_lock_t	dict_operation_lock;

 /* Dictionary system struct */
 struct dict_sys_struct{
--- a/innobase/include/fil0fil.h
+++ b/innobase/include/fil0fil.h
@ -73,6 +73,8 @@ extern fil_addr_t	fil_addr_null;
 /* File page types */
 #define FIL_PAGE_INDEX		17855
 #define FIL_PAGE_UNDO_LOG	2
+#define FIL_PAGE_INODE		3
+#define FIL_PAGE_IBUF_FREE_LIST	4

 /* Space types */
 #define FIL_TABLESPACE 		501
--- a/innobase/include/lock0lock.h
+++ b/innobase/include/lock0lock.h
@ -292,6 +292,27 @@ lock_sec_rec_modify_check_and_lock(
 	dict_index_t*	index,	/* in: secondary index */
 	que_thr_t*	thr);	/* in: query thread */
 /*************************************************************************
+Like the counterpart for a clustered index below, but now we read a
+secondary index record. */
+
+ulint
+lock_sec_rec_read_check_and_lock(
+/*=============================*/
+				/* out: DB_SUCCESS, DB_LOCK_WAIT,
+				DB_DEADLOCK, or DB_QUE_THR_SUSPENDED */
+	ulint		flags,	/* in: if BTR_NO_LOCKING_FLAG bit is set,
+				does nothing */
+	rec_t*		rec,	/* in: user record or page supremum record
+				which should be read or passed over by a read
+				cursor */
+	dict_index_t*	index,	/* in: secondary index */
+	ulint		mode,	/* in: mode of the lock which the read cursor
+				should set on records: LOCK_S or LOCK_X; the
+				latter is possible in SELECT FOR UPDATE */
+	ulint		gap_mode,/* in: LOCK_ORDINARY, LOCK_GAP, or
+				LOCK_REC_NOT_GAP */
+	que_thr_t*	thr);	/* in: query thread */
+/*************************************************************************
 Checks if locks of other transactions prevent an immediate read, or passing
 over by a read cursor, of a clustered index record. If they do, first tests
 if the query thread should anyway be suspended for some reason; if not, then
@ -313,25 +334,8 @@ lock_clust_rec_read_check_and_lock(
 	ulint		mode,	/* in: mode of the lock which the read cursor
 				should set on records: LOCK_S or LOCK_X; the
 				latter is possible in SELECT FOR UPDATE */
-	que_thr_t*	thr);	/* in: query thread */
-/*************************************************************************
-Like the counterpart for a clustered index above, but now we read a
-secondary index record. */
-
-ulint
-lock_sec_rec_read_check_and_lock(
-/*=============================*/
-				/* out: DB_SUCCESS, DB_LOCK_WAIT,
-				DB_DEADLOCK, or DB_QUE_THR_SUSPENDED */
-	ulint		flags,	/* in: if BTR_NO_LOCKING_FLAG bit is set,
-				does nothing */
-	rec_t*		rec,	/* in: user record or page supremum record
-				which should be read or passed over by a read
-				cursor */
-	dict_index_t*	index,	/* in: secondary index */
-	ulint		mode,	/* in: mode of the lock which the read cursor
-				should set on records: LOCK_S or LOCK_X; the
-				latter is possible in SELECT FOR UPDATE */
+	ulint		gap_mode,/* in: LOCK_ORDINARY, LOCK_GAP, or
+				LOCK_REC_NOT_GAP */
 	que_thr_t*	thr);	/* in: query thread */
 /*************************************************************************
 Checks that a record is seen in a consistent read. */
@ -509,6 +513,7 @@ lock_validate(void);
 extern lock_sys_t*	lock_sys;

 /* Lock modes and types */
+/* Basic modes */
 #define	LOCK_NONE	0	/* this flag is used elsewhere to note
 				consistent read */
 #define	LOCK_IS		2	/* intention shared */
@ -519,15 +524,20 @@ extern lock_sys_t*	lock_sys;
 				in an exclusive mode */
 #define LOCK_MODE_MASK	0xF	/* mask used to extract mode from the
 				type_mode field in a lock */
+/* Lock types */
 #define LOCK_TABLE	16	/* these type values should be so high that */
 #define	LOCK_REC	32	/* they can be ORed to the lock mode */
 #define LOCK_TYPE_MASK	0xF0	/* mask used to extract lock type from the
 				type_mode field in a lock */
+/* Waiting lock flag */
 #define LOCK_WAIT	256	/* this wait bit should be so high that
 				it can be ORed to the lock mode and type;
 				when this bit is set, it means that the
 				lock has not yet been granted, it is just
 				waiting for its turn in the wait queue */
+/* Precise modes */
+#define LOCK_ORDINARY	0	/* this flag denotes an ordinary next-key lock
+				in contrast to LOCK_GAP or LOCK_REC_NOT_GAP */ 
 #define LOCK_GAP	512	/* this gap bit should be so high that
 				it can be ORed to the other flags;
 				when this bit is set, it means that the
@ -537,7 +547,15 @@ extern lock_sys_t*	lock_sys;
 				the bit is set; locks of this type are created
 				when records are removed from the index chain
 				of records */
-#define LOCK_INSERT_INTENTION 1024 /* this bit is set when we place a waiting
+#define LOCK_REC_NOT_GAP 1024 	/* this bit means that the lock is only on
+				the index record and does NOT block inserts
+				to the gap before the index record; this is
+				used in the case when we retrieve a record
+				with a unique key, and is also used in
+				locking plain SELECTs (not part of UPDATE
+				or DELETE) when the user has set the READ
+				COMMITTED isolation level */
+#define LOCK_INSERT_INTENTION 2048 /* this bit is set when we place a waiting
 				gap type record lock request in order to let
 				an insert of an index record to wait until
 				there are no conflicting locks by other
--- a/innobase/include/log0recv.h
+++ b/innobase/include/log0recv.h
@ -334,6 +334,7 @@ extern ibool		recv_no_ibuf_operations;
 extern ibool		recv_needed_recovery;

 extern ibool		recv_is_making_a_backup;
+extern ulint		recv_max_parsed_page_no;

 /* Size of the parsing buffer; it must accommodate RECV_SCAN_SIZE many
 times! */ 
--- a/innobase/include/os0file.h
+++ b/innobase/include/os0file.h
@ -111,6 +111,7 @@ log. */
 #define OS_WIN31     1
 #define OS_WIN95     2	
 #define OS_WINNT     3
+#define OS_WIN2000   4

 extern ulint	os_n_file_reads;
 extern ulint	os_n_file_writes;
@ -122,7 +123,7 @@ Gets the operating system version. Currently works only on Windows. */
 ulint
 os_get_os_version(void);
 /*===================*/
-                  /* out: OS_WIN95, OS_WIN31, OS_WINNT (2000 == NT) */
+                  /* out: OS_WIN95, OS_WIN31, OS_WINNT, or OS_WIN2000 */
 /********************************************************************
 Creates the seek mutexes used in positioned reads and writes. */

--- a/innobase/include/os0proc.h
+++ b/innobase/include/os0proc.h
@ -15,6 +15,15 @@ Created 9/30/1995 Heikki Tuuri
 typedef void*			os_process_t;
 typedef unsigned long int	os_process_id_t;

+/********************************************************************
+Converts the current process id to a number. It is not guaranteed that the
+number is unique. In Linux returns the 'process number' of the current
+thread. That number is the same as one sees in 'top', for example. In Linux
+the thread id is not the same as one sees in 'top'. */
+
+ulint
+os_proc_get_number(void);
+/*====================*/
 /********************************************************************
 Allocates non-cacheable memory. */

--- a/innobase/include/os0thread.h
+++ b/innobase/include/os0thread.h
@ -16,11 +16,8 @@ Created 9/8/1995 Heikki Tuuri
 this is also the size of the wait slot array for MySQL threads which
 can wait inside InnoDB */
 #ifdef __WIN__
-/* Windows 95/98/ME seemed to have difficulties creating the all
-the event semaphores for the wait array slots. If the computer had
-<= 64 MB memory, InnoDB startup could take minutes or even crash.
-That is why we set this to only 1000 in Windows. */
-
+/* Create less event semaphores because Win 98/ME had difficult creating
+40000 event semaphores */
 #define	OS_THREAD_MAX_N		1000
 #else
 #define	OS_THREAD_MAX_N		10000
--- a/innobase/include/page0cur.h
+++ b/innobase/include/page0cur.h
@ -26,7 +26,12 @@ Created 10/4/1994 Heikki Tuuri
 #define	PAGE_CUR_GE	2
 #define	PAGE_CUR_L	3
 #define	PAGE_CUR_LE	4
-#define	PAGE_CUR_DBG	5
+#define PAGE_CUR_LE_OR_EXTENDS 5 /* This is a search mode used in
+				 "column LIKE 'abc%' ORDER BY column DESC";
+				 we have to find strings which are <= 'abc' or
+				 which extend it */
+#define	PAGE_CUR_DBG	6
+

 extern ulint	page_cur_short_succ;

--- a/innobase/include/page0page.h
+++ b/innobase/include/page0page.h
@ -666,6 +666,16 @@ page_rec_validate(
 			/* out: TRUE if ok */
 	rec_t* 	rec);	/* in: record on the page */
 /*******************************************************************
+This function checks the consistency of an index page when we do not
+know the index. This is also resilient so that this should never crash
+even if the page is total garbage. */
+
+ibool
+page_simple_validate(
+/*=================*/
+			/* out: TRUE if ok */
+	page_t*	page);	/* in: index page */
+/*******************************************************************
 This function checks the consistency of an index page. */

 ibool
--- a/innobase/include/que0que.h
+++ b/innobase/include/que0que.h
@ -117,6 +117,7 @@ que_thr_stop(
 /**************************************************************************
 Moves a thread from another state to the QUE_THR_RUNNING state. Increments
 the n_active_thrs counters of the query graph and transaction. */
+
 void
 que_thr_move_to_run_state_for_mysql(
 /*================================*/
@ -125,14 +126,17 @@ que_thr_move_to_run_state_for_mysql(
 /**************************************************************************
 A patch for MySQL used to 'stop' a dummy query thread used in MySQL
 select, when there is no error or lock wait. */
+
 void
 que_thr_stop_for_mysql_no_error(
 /*============================*/
 	que_thr_t*	thr,	/* in: query thread */
 	trx_t*		trx);	/* in: transaction */
 /**************************************************************************
-A patch for MySQL used to 'stop' a dummy query thread used in MySQL
-select. */
+A patch for MySQL used to 'stop' a dummy query thread used in MySQL. The
+query thread is stopped and made inactive, except in the case where
+it was put to the lock wait state in lock0lock.c, but the lock has already
+been granted or the transaction chosen as a victim in deadlock resolution. */

 void
 que_thr_stop_for_mysql(
--- a/innobase/include/read0read.h
+++ b/innobase/include/read0read.h
@ -45,6 +45,14 @@ read_view_close(
 /*============*/
 	read_view_t*	view);	/* in: read view */
 /*************************************************************************
+Closes a consistent read view for MySQL. This function is called at an SQL
+statement end if the trx isolation level is <= TRX_ISO_READ_COMMITTED. */
+
+void
+read_view_close_for_mysql(
+/*======================*/
+	trx_t*	trx);	/* in: trx which has a read view */
+/*************************************************************************
 Checks if a read view sees the specified transaction. */
 UNIV_INLINE
 ibool
--- a/innobase/include/rem0rec.h
+++ b/innobase/include/rem0rec.h
@ -148,12 +148,22 @@ data field in the record. */
 byte*
 rec_get_nth_field(
 /*==============*/
- 			/* out: pointer to the field, NULL if SQL null */
+ 			/* out: pointer to the field */
 	rec_t*	rec, 	/* in: record */
 	ulint	n,	/* in: index of the field */
 	ulint*	len);	/* out: length of the field; UNIV_SQL_NULL 
 			if SQL null */
 /****************************************************************
+Return field length or UNIV_SQL_NULL. */
+UNIV_INLINE
+ulint
+rec_get_nth_field_len(
+/*==================*/
+ 			/* out: length of the field; UNIV_SQL_NULL if SQL
+			null */
+ 	rec_t*	rec, 	/* in: record */
+ 	ulint	n);	/* in: index of the field */
+/****************************************************************
 Gets the physical size of a field. Also an SQL null may have a field of
 size > 0, if the data type is of a fixed size. */
 UNIV_INLINE
--- a/innobase/include/rem0rec.ic
+++ b/innobase/include/rem0rec.ic
@ -65,6 +65,24 @@ a field stored to another page: */

 #define REC_2BYTE_EXTERN_MASK	0x4000

+/****************************************************************
+Return field length or UNIV_SQL_NULL. */
+UNIV_INLINE
+ulint
+rec_get_nth_field_len(
+/*==================*/
+ 			/* out: length of the field; UNIV_SQL_NULL if SQL
+			null */
+ 	rec_t*	rec, 	/* in: record */
+ 	ulint	n)	/* in: index of the field */
+{
+	ulint	len;
+
+	rec_get_nth_field(rec, n, &len);
+
+	return(len);
+}
+
 /***************************************************************
 Sets the value of the ith field SQL null bit. */

--- a/innobase/include/row0mysql.h
+++ b/innobase/include/row0mysql.h
@ -230,18 +230,35 @@ row_update_cascade_for_mysql(
 				or set null operation */
 	dict_table_t*	table);	/* in: table where we do the operation */
 /*************************************************************************
-Locks the data dictionary exclusively for performing a table create
-operation. */
+Locks the data dictionary exclusively for performing a table create or other
+data dictionary modification operation. */

 void
-row_mysql_lock_data_dictionary(void);
-/*================================*/
+row_mysql_lock_data_dictionary(
+/*===========================*/
+	trx_t*	trx);	/* in: transaction */
 /*************************************************************************
-Unlocks the data dictionary exclusively lock. */
+Unlocks the data dictionary exclusive lock. */

 void
-row_mysql_unlock_data_dictionary(void);
-/*==================================*/
+row_mysql_unlock_data_dictionary(
+/*=============================*/
+	trx_t*	trx);	/* in: transaction */
+/*************************************************************************
+Locks the data dictionary in shared mode from modifications, for performing
+foreign key check, rollback, or other operation invisible to MySQL. */
+
+void
+row_mysql_freeze_data_dictionary(
+/*=============================*/
+	trx_t*	trx);	/* in: transaction */
+/*************************************************************************
+Unlocks the data dictionary shared lock. */
+
+void
+row_mysql_unfreeze_data_dictionary(
+/*===============================*/
+	trx_t*	trx);	/* in: transaction */
 /*************************************************************************
 Does a table creation operation for MySQL. If the name of the created
 table ends to characters INNODB_MONITOR, then this also starts
@ -312,9 +329,7 @@ row_drop_table_for_mysql(
 /*=====================*/
 			/* out: error code or DB_SUCCESS */
 	char*	name,	/* in: table name */
-	trx_t*	trx,		/* in: transaction handle */
-	ibool	has_dict_mutex);/* in: TRUE if the caller already owns the
-				dictionary system mutex */
+	trx_t*	trx);	/* in: transaction handle */
 /*************************************************************************
 Drops a database for MySQL. */

@ -393,6 +408,10 @@ struct row_prebuilt_struct {
 					an SQL statement: we may have to set
 					an intention lock on the table,
 					create a consistent read view etc. */
+        ibool           mysql_has_locked; /* this is set TRUE when MySQL
+			                calls external_lock on this handle
+			                with a lock flag, and set FALSE when
+			                with the F_UNLOCK flag */
 	ibool		clust_index_was_generated;
 					/* if the user did not define a
 					primary key in MySQL, then Innobase
--- a/innobase/include/srv0srv.h
+++ b/innobase/include/srv0srv.h
@ -28,6 +28,9 @@ extern os_event_t	srv_lock_timeout_thread_event;
 at a time */
 #define SRV_AUTO_EXTEND_INCREMENT   (8 * ((1024 * 1024) / UNIV_PAGE_SIZE))

+/* This is set to TRUE if the MySQL user has set it in MySQL */
+extern ibool	srv_lower_case_table_names;
+
 /* Server parameters which are read from the initfile */

 extern char*	srv_data_home;
@ -57,8 +60,6 @@ extern ulint	srv_flush_log_at_trx_commit;

 extern byte	srv_latin1_ordering[256];/* The sort order table of the latin1
 					character set */
-extern ibool	srv_use_native_aio;		
-
 extern ulint	srv_pool_size;
 extern ulint	srv_mem_pool_size;
 extern ulint	srv_lock_table_size;
@ -70,8 +71,9 @@ extern dulint	srv_archive_recovery_limit_lsn;

 extern ulint	srv_lock_wait_timeout;

-extern char*    srv_unix_file_flush_method_str;
+extern char*    srv_file_flush_method_str;
 extern ulint    srv_unix_file_flush_method;
+extern ulint   	srv_win_file_flush_method;
 extern ulint	srv_force_recovery;
 extern ulint	srv_thread_concurrency;

@ -154,13 +156,19 @@ typedef struct srv_sys_struct	srv_sys_t;
 /* The server system */
 extern srv_sys_t*	srv_sys;

-/* Alternatives for the field flush option in Unix; see the InnoDB manual about
+/* Alternatives for the file flush option in Unix; see the InnoDB manual about
 what these mean */
-#define SRV_UNIX_FDATASYNC   1
+#define SRV_UNIX_FDATASYNC   1	/* This is the default; it is currently mapped
+				to a call of fsync() because fdatasync()
+				seemed to corrupt files in Linux and Solaris */
 #define SRV_UNIX_O_DSYNC     2
 #define SRV_UNIX_LITTLESYNC  3
 #define SRV_UNIX_NOSYNC      4

+/* Alternatives for file i/o in Windows */
+#define SRV_WIN_IO_NORMAL		1
+#define SRV_WIN_IO_UNBUFFERED		2	/* This is the default */
+
 /* Alternatives for srv_force_recovery. Non-zero values are intended
 to help the user get a damaged database up so that he can dump intact
 tables and rows with SELECT INTO OUTFILE. The database must not otherwise
@ -311,15 +319,17 @@ srv_conc_exit_innodb(
 	trx_t*	trx);	/* in: transaction object associated with the
 			thread */
 /*******************************************************************
-Puts a MySQL OS thread to wait for a lock to be released. */
+Puts a MySQL OS thread to wait for a lock to be released. If an error
+occurs during the wait trx->error_state associated with thr is
+!= DB_SUCCESS when we return. DB_LOCK_WAIT_TIMEOUT and DB_DEADLOCK
+are possible errors. DB_DEADLOCK is returned if selective deadlock
+resolution chose this transaction as a victim. */

-ibool
+void
 srv_suspend_mysql_thread(
 /*=====================*/
-				/* out: TRUE if the lock wait timeout was
-				exceeded */
-	que_thr_t*	thr);	/* in: query thread associated with
-				the MySQL OS thread */
+	que_thr_t*	thr);	/* in: query thread associated with the MySQL
+				OS thread */
 /************************************************************************
 Releases a MySQL OS thread waiting for a lock to be released, if the
 thread is already suspended. */
@ -407,3 +417,4 @@ struct srv_sys_struct{
 extern ulint	srv_n_threads_active[];

 #endif
+
--- a/innobase/include/srv0start.h
+++ b/innobase/include/srv0start.h
@ -79,6 +79,8 @@ innobase_shutdown_for_mysql(void);
 /*=============================*/
 				/* out: DB_SUCCESS or error code */

+extern  ulint   srv_sizeof_trx_t_in_ha_innodb_cc;
+
 extern	ibool	srv_startup_is_before_trx_rollback_phase;
 extern	ibool	srv_is_being_shut_down;

--- a/innobase/include/sync0rw.h
+++ b/innobase/include/sync0rw.h
@ -335,7 +335,8 @@ ibool
 rw_lock_own(
 /*========*/
 	rw_lock_t*	lock,		/* in: rw-lock */
-	ulint		lock_type);	/* in: lock type */
+	ulint		lock_type);	/* in: lock type: RW_LOCK_SHARED,
+					RW_LOCK_EX */
 /**********************************************************************
 Checks if somebody has locked the rw-lock in the specified mode. */

--- a/innobase/include/sync0sync.h
+++ b/innobase/include/sync0sync.h
@ -371,10 +371,12 @@ or row lock! */
 #define SYNC_NO_ORDER_CHECK	3000	/* this can be used to suppress
 					latching order checking */
 #define	SYNC_LEVEL_NONE		2000	/* default: level not defined */
-#define	SYNC_FOREIGN_KEY_CHECK	1001
+#define	SYNC_DICT_OPERATION	1001	/* table create, drop, etc. reserve
+					this in X-mode, implicit or backround
+					operations purge, rollback, foreign
+					key checks reserve this in S-mode */
 #define SYNC_DICT		1000
 #define SYNC_DICT_AUTOINC_MUTEX	999
-#define	SYNC_PURGE_IS_RUNNING	997
 #define SYNC_DICT_HEADER	995
 #define SYNC_IBUF_HEADER	914
 #define SYNC_IBUF_PESS_INSERT_MUTEX 912
--- a/innobase/include/trx0purge.h
+++ b/innobase/include/trx0purge.h
@ -111,9 +111,6 @@ struct trx_purge_struct{
 					of the trx system and it never ends */
 	que_t*		query;		/* The query graph which will do the
 					parallelized purge operation */
-	rw_lock_t	purge_is_running;/* Purge operation set an x-latch here
-					while it is accessing a table: this
-					prevents dropping of the table */
 	rw_lock_t	latch;		/* The latch protecting the purge view.
 					A purge operation must acquire an
 					x-latch here for the instant at which
--- a/innobase/include/trx0trx.h
+++ b/innobase/include/trx0trx.h
@ -327,6 +327,7 @@ struct trx_struct{
        time_t          start_time;     /* time the trx object was created
                                        or the state last time became
                                        TRX_ACTIVE */
+	ulint		isolation_level;/* TRX_ISO_REPEATABLE_READ, ... */
 	ibool		check_foreigns;	/* normally TRUE, but if the user
 					wants to suppress foreign key checks,
 					(in table imports, for example) we
@ -350,6 +351,9 @@ struct trx_struct{
 	/*------------------------------*/
        void*           mysql_thd;      /* MySQL thread handle corresponding
                                        to this trx, or NULL */
+	char**		mysql_query_str;/* pointer to the field in mysqld_thd
+					which contains the pointer to the
+					current SQL query string */
 	char*		mysql_log_file_name;
 					/* if MySQL binlog is used, this field
 					contains a pointer to the latest file
@ -371,6 +375,9 @@ struct trx_struct{
 					replication has processed */
 	os_thread_id_t	mysql_thread_id;/* id of the MySQL thread associated
 					with this transaction object */
+	ulint		mysql_process_no;/* since in Linux, 'top' reports
+					process id's and not thread id's, we
+					store the process number too */
 	/*------------------------------*/
 	ulint		n_mysql_tables_in_use; /* number of Innobase tables
 					used in the processing of the current
@ -379,9 +386,10 @@ struct trx_struct{
                                        /* how many tables the current SQL
 					statement uses, except those
 					in consistent read */
-	ibool		has_dict_foreign_key_check_lock;
-					/* TRUE if the trx currently holds
-					an s-lock on dict_foreign_... */
+	ibool		dict_operation_lock_mode;
+					/* 0, RW_S_LATCH, or RW_X_LATCH:
+					the latch mode trx currently holds
+					on dict_operation_lock */
        ibool           has_search_latch;
 			                /* TRUE if this trx has latched the
 			                search system latch in S-mode */
@ -420,36 +428,11 @@ struct trx_struct{
 			mysql_trx_list;	/* list of transactions created for
 					MySQL */
 	/*------------------------------*/
-	mutex_t		undo_mutex;	/* mutex protecting the fields in this
-					section (down to undo_no_arr), EXCEPT
-					last_sql_stat_start, which can be
-					accessed only when we know that there
-					cannot be any activity in the undo
-					logs! */
-	dulint		undo_no;	/* next undo log record number to
-					assign */
-	trx_savept_t	last_sql_stat_start;
-					/* undo_no when the last sql statement
-					was started: in case of an error, trx
-					is rolled back down to this undo
-					number; see note at undo_mutex! */
-	trx_rseg_t*	rseg;		/* rollback segment assigned to the
-					transaction, or NULL if not assigned
-					yet */
-	trx_undo_t*	insert_undo;	/* pointer to the insert undo log, or 
-					NULL if no inserts performed yet */
-	trx_undo_t* 	update_undo;	/* pointer to the update undo log, or
-					NULL if no update performed yet */
-	dulint		roll_limit;	/* least undo number to undo during
-					a rollback */
-	ulint		pages_undone;	/* number of undo log pages undone
-					since the last undo log truncation */
-	trx_undo_arr_t*	undo_no_arr;	/* array of undo numbers of undo log
-					records which are currently processed
-					by a rollback operation */
-	/*------------------------------*/
 	ulint		error_state;	/* 0 if no error, otherwise error
-					number */
+					number; NOTE That ONLY the thread
+					doing the transaction is allowed to
+					set this field: this is NOT protected
+					by the kernel mutex */
 	void*		error_info;	/* if the error number indicates a
 					duplicate key error, a pointer to
 					the problematic index is stored here */
@ -486,6 +469,12 @@ struct trx_struct{
 					TRX_QUE_LOCK_WAIT, this points to
 					the lock request, otherwise this is
 					NULL */
+	ibool		was_chosen_as_deadlock_victim;
+					/* when the transaction decides to wait
+					for a lock, this it sets this to FALSE;
+					if another transaction chooses this
+					transaction as a victim in deadlock
+					resolution, it sets this to TRUE */
 	time_t          wait_started;   /* lock wait started at this time */
 	UT_LIST_BASE_NODE_T(que_thr_t)
 			wait_thrs;	/* query threads belonging to this
@ -501,6 +490,34 @@ struct trx_struct{
 	/*------------------------------*/
 	mem_heap_t*	read_view_heap;	/* memory heap for the read view */
 	read_view_t*	read_view;	/* consistent read view or NULL */
+	/*------------------------------*/
+	mutex_t		undo_mutex;	/* mutex protecting the fields in this
+					section (down to undo_no_arr), EXCEPT
+					last_sql_stat_start, which can be
+					accessed only when we know that there
+					cannot be any activity in the undo
+					logs! */
+	dulint		undo_no;	/* next undo log record number to
+					assign */
+	trx_savept_t	last_sql_stat_start;
+					/* undo_no when the last sql statement
+					was started: in case of an error, trx
+					is rolled back down to this undo
+					number; see note at undo_mutex! */
+	trx_rseg_t*	rseg;		/* rollback segment assigned to the
+					transaction, or NULL if not assigned
+					yet */
+	trx_undo_t*	insert_undo;	/* pointer to the insert undo log, or 
+					NULL if no inserts performed yet */
+	trx_undo_t* 	update_undo;	/* pointer to the update undo log, or
+					NULL if no update performed yet */
+	dulint		roll_limit;	/* least undo number to undo during
+					a rollback */
+	ulint		pages_undone;	/* number of undo log pages undone
+					since the last undo log truncation */
+	trx_undo_arr_t*	undo_no_arr;	/* array of undo numbers of undo log
+					records which are currently processed
+					by a rollback operation */
 };

 #define TRX_MAX_N_THREADS	32	/* maximum number of concurrent
@ -523,6 +540,41 @@ struct trx_struct{
 #define TRX_QUE_ROLLING_BACK	3	/* transaction is rolling back */
 #define TRX_QUE_COMMITTING	4	/* transaction is committing */

+/* Transaction isolation levels */
+#define TRX_ISO_READ_UNCOMMITTED	1	/* dirty read: non-locking
+						SELECTs are performed so that
+						we do not look at a possible
+						earlier version of a record;
+						thus they are not 'consistent'
+						reads under this isolation
+						level; otherwise like level
+						2 */
+
+#define TRX_ISO_READ_COMMITTED		2	/* somewhat Oracle-like
+						isolation, except that in
+						range UPDATE and DELETE we
+						must block phantom rows
+						with next-key locks;
+						SELECT ... FOR UPDATE and ...
+						LOCK IN SHARE MODE only lock
+						the index records, NOT the
+						gaps before them, and thus
+						allow free inserting;
+						each consistent read reads its
+						own snapshot */
+
+#define TRX_ISO_REPEATABLE_READ		3	/* this is the default;
+						all consistent reads in the
+						same trx read the same
+						snapshot;
+						full next-key locking used
+						in locking reads to block
+						insertions into gaps */
+
+#define TRX_ISO_SERIALIZABLE		4	/* all plain SELECTs are
+						converted to LOCK IN SHARE
+						MODE reads */
+
 /* Types of a trx signal */
 #define TRX_SIG_NO_SIGNAL		100
 #define TRX_SIG_TOTAL_ROLLBACK		1
--- a/innobase/lock/lock0lock.c
+++ b/innobase/lock/lock0lock.c
@ -70,6 +70,11 @@ A waiting record lock can also be of the gap type. A waiting lock request
 can be granted when there is no conflicting mode lock request by another
 transaction ahead of it in the explicit lock queue.

+In version 4.0.5 we added yet another explicit lock type: LOCK_REC_NOT_GAP.
+It only locks the record it is placed on, not the gap before the record.
+This lock type is necessary to emulate an Oracle-like READ COMMITTED isolation
+level.
+
 -------------------------------------------------------------------------
 RULE 1: If there is an implicit x-lock on a record, and there are non-gap
 -------
@ -294,7 +299,9 @@ struct lock_struct{
 	UT_LIST_NODE_T(lock_t)		
 			trx_locks;	/* list of the locks of the
 					transaction */
-	ulint		type_mode;	/* lock type, mode, gap flag, and
+	ulint		type_mode;	/* lock type, mode, LOCK_GAP or
+					LOCK_REC_NOT_GAP,
+					LOCK_INSERT_INTENTION,
 					wait flag, ORed */
 	hash_node_t	hash;		/* hash chain node for a record lock */
 	dict_index_t*	index;		/* index for a record lock */
@ -309,6 +316,10 @@ Monitor will then fetch it and print */
 ibool	lock_deadlock_found = FALSE;
 char*	lock_latest_err_buf;		/* We allocate 5000 bytes for this */

+/* Flags for recursive deadlock search */
+#define LOCK_VICTIM_IS_START	1
+#define LOCK_VICTIM_IS_OTHER	2
+
 /************************************************************************
 Checks if a lock request results in a deadlock. */
 static
@ -700,23 +711,23 @@ lock_rec_get_gap(
 }

 /*************************************************************************
-Sets the gap flag of a record lock. */
+Gets the LOCK_REC_NOT_GAP flag of a record lock. */
 UNIV_INLINE
-void
-lock_rec_set_gap(
-/*=============*/
-	lock_t*	lock,	/* in: record lock */
-	ibool	val)	/* in: value to set: TRUE or FALSE */
+ibool
+lock_rec_get_rec_not_gap(
+/*=====================*/
+			/* out: TRUE if LOCK_REC_NOT_GAP flag set */
+	lock_t*	lock)	/* in: record lock */
 {
 	ut_ad(lock);
-	ut_ad((val == TRUE) || (val == FALSE));
 	ut_ad(lock_get_type(lock) == LOCK_REC);

-	if (val) {
- 		lock->type_mode = lock->type_mode | LOCK_GAP;
-	} else {
-		lock->type_mode = lock->type_mode & ~LOCK_GAP;
+	if (lock->type_mode & LOCK_REC_NOT_GAP) {
+
+		return(TRUE);
 	}
+
+	return(FALSE);
 }

 /*************************************************************************
@ -739,26 +750,6 @@ lock_rec_get_insert_intention(
 	return(FALSE);
 }

-/*************************************************************************
-Sets the waiting insert flag of a record lock. */
-UNIV_INLINE
-void
-lock_rec_set_insert_intention(
-/*==========================*/
-	lock_t*	lock,	/* in: record lock */
-	ibool	val)	/* in: value to set: TRUE or FALSE */
-{
-	ut_ad(lock);
-	ut_ad((val == TRUE) || (val == FALSE));
-	ut_ad(lock_get_type(lock) == LOCK_REC);
-
-	if (val) {
- 		lock->type_mode = lock->type_mode | LOCK_INSERT_INTENTION;
-	} else {
-		lock->type_mode = lock->type_mode & ~LOCK_INSERT_INTENTION;
-	}
-}
-
 /*************************************************************************
 Calculates if lock mode 1 is stronger or equal to lock mode 2. */
 UNIV_INLINE
@ -848,47 +839,52 @@ lock_rec_has_to_wait(
 			/* out: TRUE if new lock has to wait for lock2 to be
 			removed */
 	trx_t*	trx,	/* in: trx of new lock */
-	ulint	mode,	/* in: LOCK_S or LOCK_X */
-	ulint	gap,	/* in: LOCK_GAP or 0 */
-	ulint	insert_intention,
-			/* in: LOCK_INSERT_INTENTION or 0 */
+	ulint	type_mode,/* in: precise mode of the new lock to set:
+			LOCK_S or LOCK_X, possibly ORed to
+			LOCK_GAP or LOCK_REC_NOT_GAP, LOCK_INSERT_INTENTION */
 	lock_t*	lock2)	/* in: another record lock; NOTE that it is assumed
 			that this has a lock bit set on the same record as
-			in lock1 */
+			in the new lock we are setting */
 {
 	ut_ad(trx && lock2);
 	ut_ad(lock_get_type(lock2) == LOCK_REC);
-	ut_ad(mode == LOCK_S || mode == LOCK_X);
-	ut_ad(gap == LOCK_GAP || gap == 0);
-	ut_ad(insert_intention == LOCK_INSERT_INTENTION
-	      				|| insert_intention == 0);

-	if (trx != lock2->trx && !lock_mode_compatible(mode,
+	if (trx != lock2->trx
+	    && !lock_mode_compatible(LOCK_MODE_MASK & type_mode,
 				     		lock_get_mode(lock2))) {

-		/* We have somewhat complex rules when gap type
-		record locks cause waits */
+		/* We have somewhat complex rules when gap type record locks
+		cause waits */

-		if (!gap && lock_rec_get_insert_intention(lock2)) {
-
-			/* Request of a full next-key record does not
-			need to wait for an insert intention lock to be
-			removed. This is ok since our rules allow conflicting
-			locks on gaps. This eliminates a spurious deadlock
-			caused by a next-key lock waiting for an insert
-			intention lock; when the insert intention lock was
-			granted, the insert deadlocked on the waiting
-			next-key lock. */
+		if ((type_mode & LOCK_REC_NOT_GAP)
+						&& lock_rec_get_gap(lock2)) {
+			/* Lock on just the record does not need to wait for
+			a gap type lock */

 			return(FALSE);
 		}

-		if (insert_intention && lock_rec_get_insert_intention(lock2)) {
+		if ((type_mode & LOCK_GAP)
+					&& lock_rec_get_rec_not_gap(lock2)) {
 		
-			/* An insert intention is not disturbed by another
-			insert intention; this removes a spurious deadlock
-			caused by inserts which had to wait for a next-key
-			lock to be removed */
+			/* Lock on gap does not need to wait for
+			a LOCK_REC_NOT_GAP type lock */
+
+			return(FALSE);
+		}
+
+		if (lock_rec_get_insert_intention(lock2)) {
+
+			/* No lock request needs to wait for an insert
+			intention lock to be removed. This is ok since our
+			rules allow conflicting locks on gaps. This eliminates
+			a spurious deadlock caused by a next-key lock waiting
+			for an insert intention lock; when the insert
+			intention lock was granted, the insert deadlocked on
+			the waiting next-key lock.
+
+			Also, insert intention locks do not disturb each
+			other. */
 				
 			return(FALSE);
 		}
@ -921,10 +917,7 @@ lock_has_to_wait(
 			ut_ad(lock_get_type(lock2) == LOCK_REC);
 				
 			return(lock_rec_has_to_wait(lock1->trx,
-				lock_get_mode(lock1),
-				lock_rec_get_gap(lock1),
-				lock_rec_get_insert_intention(lock1),
-				lock2));
+						lock1->type_mode, lock2));
 		}

 		return(TRUE);
@ -1386,32 +1379,41 @@ lock_table_has(
 /*============= FUNCTIONS FOR ANALYZING RECORD LOCK QUEUE ================*/

 /*************************************************************************
-Checks if a transaction has a GRANTED explicit lock on rec, where the gap
-flag or the insert intention flag is not set, stronger or equal to mode.
-Note that locks on the supremum of a page are a special case here, since
-they are always gap type locks, even if the gap flag is not set in them. */
+Checks if a transaction has a GRANTED explicit lock on rec stronger or equal
+to precise_mode. */
 UNIV_INLINE
 lock_t*
 lock_rec_has_expl(
 /*==============*/
 			/* out: lock or NULL */
-	ulint	mode,	/* in: lock mode */
+	ulint	precise_mode,/* in: LOCK_S or LOCK_X possibly ORed to
+			LOCK_GAP or LOCK_REC_NOT_GAP,
+			for a supremum record we regard this always a gap
+			type request */
 	rec_t*	rec,	/* in: record */
 	trx_t*	trx)	/* in: transaction */
 {
 	lock_t*	lock;

 	ut_ad(mutex_own(&kernel_mutex));
-	ut_ad((mode == LOCK_X) || (mode == LOCK_S));
+	ut_ad((precise_mode & LOCK_MODE_MASK) == LOCK_S
+	      || (precise_mode & LOCK_MODE_MASK) == LOCK_X);
+	ut_ad(!(precise_mode & LOCK_INSERT_INTENTION));
 	
 	lock = lock_rec_get_first(rec);

 	while (lock) {
 		if (lock->trx == trx
-		    && lock_mode_stronger_or_eq(lock_get_mode(lock), mode)
+		    && lock_mode_stronger_or_eq(lock_get_mode(lock),
+		    				precise_mode & LOCK_MODE_MASK)
 		    && !lock_get_wait(lock)
-		    && !lock_rec_get_insert_intention(lock)
-		    && !lock_rec_get_gap(lock)) {
+		    && (!lock_rec_get_rec_not_gap(lock)
+		    		|| (precise_mode & LOCK_REC_NOT_GAP)
+		    		|| page_rec_is_supremum(rec))
+		    && (!lock_rec_get_gap(lock)
+				|| (precise_mode & LOCK_GAP)
+				|| page_rec_is_supremum(rec))
+		    && (!lock_rec_get_insert_intention(lock))) {

 		    	return(lock);
 		}
@ -1429,7 +1431,7 @@ lock_t*
 lock_rec_other_has_expl_req(
 /*========================*/
 			/* out: lock or NULL */
-	ulint	mode,	/* in: lock mode */
+	ulint	mode,	/* in: LOCK_S or LOCK_X */
 	ulint	gap,	/* in: LOCK_GAP if also gap locks are taken
 			into account, or 0 if not */
 	ulint	wait,	/* in: LOCK_WAIT if also waiting locks are
@ -1471,27 +1473,21 @@ lock_t*
 lock_rec_other_has_conflicting(
 /*===========================*/
 			/* out: lock or NULL */
-	ulint	mode,	/* in: lock mode of the lock we are going to reserve */
-	ulint	gap,	/* in: LOCK_GAP if we are going to reserve a gap type
-			lock, else 0 */
-	ulint	insert_intention,
-			/* in: LOCK_INSERT_INTENTION if we are going to
-			reserve an insert intention lock */
+	ulint	mode,	/* in: LOCK_S or LOCK_X,
+			possibly ORed to LOCK_GAP or LOC_REC_NOT_GAP,
+			LOCK_INSERT_INTENTION */
 	rec_t*	rec,	/* in: record to look at */	
 	trx_t*	trx)	/* in: our transaction */
 {
 	lock_t*	lock;
 	
 	ut_ad(mutex_own(&kernel_mutex));
-	ut_ad(mode == LOCK_X || mode == LOCK_S);
-	ut_ad(gap == 0 || gap == LOCK_GAP);
-	ut_ad(insert_intention == LOCK_INSERT_INTENTION
-						|| insert_intention == 0);
+
 	lock = lock_rec_get_first(rec);

 	while (lock) {
-		if (lock_rec_has_to_wait(trx, mode, gap, insert_intention,
-								lock)) {
+		if (lock_rec_has_to_wait(trx, mode, lock)) {
+
 			return(lock);
 		}
 		
@ -1607,14 +1603,14 @@ lock_rec_create(
 	page_no	= buf_frame_get_page_no(page);
 	heap_no = rec_get_heap_no(rec);

-	/* If rec is the supremum record, then we reset the gap bit, as
-	all locks on the supremum are automatically of the gap type, and
-	we try to avoid unnecessary memory consumption of a new record lock
-	struct for a gap type lock */
+	/* If rec is the supremum record, then we reset the gap and
+	LOCK_REC_NOT_GAP bits, as all locks on the supremum are
+	automatically of the gap type */

 	if (rec == page_get_supremum_rec(page)) {
+		ut_ad(!(type_mode & LOCK_REC_NOT_GAP));

-		type_mode = type_mode & ~LOCK_GAP;
+		type_mode = type_mode & ~(LOCK_GAP | LOCK_REC_NOT_GAP);
 	}

 	/* Make lock bitmap bigger by a safety margin */
@ -1666,10 +1662,14 @@ ulint
 lock_rec_enqueue_waiting(
 /*=====================*/
 				/* out: DB_LOCK_WAIT, DB_DEADLOCK, or
-				DB_QUE_THR_SUSPENDED */
+				DB_QUE_THR_SUSPENDED, or DB_SUCCESS;
+				DB_SUCCESS means that there was a deadlock,
+				but another transaction was chosen as a
+				victim, and we got the lock immediately:
+				no need to wait then */
 	ulint		type_mode,/* in: lock mode this transaction is
-				requesting: LOCK_S or LOCK_X, ORed with
-				LOCK_GAP if a gap lock is requested, ORed
+				requesting: LOCK_S or LOCK_X, possibly ORed
+				with LOCK_GAP or LOCK_REC_NOT_GAP, ORed
 				with LOCK_INSERT_INTENTION if this waiting
 				lock request is set when performing an
 				insert of an index record */
@ -1718,7 +1718,16 @@ index->table_name);
 		return(DB_DEADLOCK);
 	}

+	/* If there was a deadlock but we chose another transaction as a
+	victim, it is possible that we already have the lock now granted! */
+
+	if (trx->wait_lock == NULL) {
+
+		return(DB_SUCCESS);
+	}
+
 	trx->que_state = TRX_QUE_LOCK_WAIT;
+	trx->was_chosen_as_deadlock_victim = FALSE;
 	trx->wait_started = time(NULL);

 	ut_a(que_thr_stop(thr));
@ -1744,8 +1753,8 @@ lock_rec_add_to_queue(
 /*==================*/
 				/* out: lock where the bit was set, NULL if out
 				of memory */
-	ulint		type_mode,/* in: lock mode, wait, and gap flags; type
-				is ignored and replaced by LOCK_REC */
+	ulint		type_mode,/* in: lock mode, wait, gap etc. flags;
+				type is ignored and replaced by LOCK_REC */
 	rec_t*		rec,	/* in: record on page */
 	dict_index_t*	index,	/* in: index of record */
 	trx_t*		trx)	/* in: transaction */
@ -1759,12 +1768,11 @@ lock_rec_add_to_queue(
 	ut_ad(mutex_own(&kernel_mutex));
 	ut_ad((type_mode & (LOCK_WAIT | LOCK_GAP))
 	      || ((type_mode & LOCK_MODE_MASK) != LOCK_S)
-	      || !lock_rec_other_has_expl_req(LOCK_X, 0, LOCK_WAIT,
-						rec, trx));
+	      || !lock_rec_other_has_expl_req(LOCK_X, 0, LOCK_WAIT, rec, trx));
 	ut_ad((type_mode & (LOCK_WAIT | LOCK_GAP))
 	      || ((type_mode & LOCK_MODE_MASK) != LOCK_X)
-	      || !lock_rec_other_has_expl_req(LOCK_S, 0, LOCK_WAIT,
-						rec, trx));
+	      || !lock_rec_other_has_expl_req(LOCK_S, 0, LOCK_WAIT, rec, trx));
+
 	type_mode = type_mode | LOCK_REC;

 	page = buf_frame_align(rec);
@ -1775,12 +1783,15 @@ lock_rec_add_to_queue(
 	struct for a gap type lock */

 	if (rec == page_get_supremum_rec(page)) {
+		ut_ad(!(type_mode & LOCK_REC_NOT_GAP));

-		type_mode = type_mode & ~LOCK_GAP;
+		/* There should never be LOCK_REC_NOT_GAP on a supremum
+		record, but let us play safe */
+		
+		type_mode = type_mode & ~(LOCK_GAP | LOCK_REC_NOT_GAP);
 	}

-	/* Look for a waiting lock request on the same record, or for a
-	similar record lock on the same page */
+	/* Look for a waiting lock request on the same record or on a gap */

 	heap_no = rec_get_heap_no(rec);
 	lock = lock_rec_get_first_on_page(rec);
@ -1795,6 +1806,9 @@ lock_rec_add_to_queue(
 		lock = lock_rec_get_next_on_page(lock);
 	}

+	/* Look for a similar record lock on the same page: if one is found
+	and there are no waiting lock requests, we can just set the bit */
+
 	similar_lock = lock_rec_find_similar_on_page(type_mode, rec, trx);

 	if (similar_lock && !somebody_waits && !(type_mode & LOCK_WAIT)) {
@ -1822,7 +1836,8 @@ lock_rec_lock_fast(
 	ibool		impl,	/* in: if TRUE, no lock is set if no wait
 				is necessary: we assume that the caller will
 				set an implicit lock */
-	ulint		mode,	/* in: lock mode */
+	ulint		mode,	/* in: lock mode: LOCK_X or LOCK_S possibly
+				ORed to either LOCK_GAP or LOCK_REC_NOT_GAP */
 	rec_t*		rec,	/* in: record */
 	dict_index_t*	index,	/* in: index of record */
 	que_thr_t* 	thr)	/* in: query thread */
@ -1831,7 +1846,15 @@ lock_rec_lock_fast(
 	ulint	heap_no;

 	ut_ad(mutex_own(&kernel_mutex));
-	ut_ad((mode == LOCK_X) || (mode == LOCK_S));
+	ut_ad((LOCK_MODE_MASK & mode) != LOCK_S
+		|| lock_table_has(thr_get_trx(thr), index->table, LOCK_IS));
+	ut_ad((LOCK_MODE_MASK & mode) != LOCK_X
+		|| lock_table_has(thr_get_trx(thr), index->table, LOCK_IX));
+	ut_ad((LOCK_MODE_MASK & mode) == LOCK_S
+		|| (LOCK_MODE_MASK & mode) == LOCK_X);
+	ut_ad(mode - (LOCK_MODE_MASK & mode) == LOCK_GAP
+			|| mode - (LOCK_MODE_MASK & mode) == 0
+			|| mode - (LOCK_MODE_MASK & mode) == LOCK_REC_NOT_GAP);
 			
 	heap_no = rec_get_heap_no(rec);
 	
@ -1877,7 +1900,8 @@ lock_rec_lock_slow(
 	ibool		impl,	/* in: if TRUE, no lock is set if no wait is
 				necessary: we assume that the caller will set
 				an implicit lock */
-	ulint		mode,	/* in: lock mode */
+	ulint		mode,	/* in: lock mode: LOCK_X or LOCK_S possibly
+				ORed to either LOCK_GAP or LOCK_REC_NOT_GAP */
 	rec_t*		rec,	/* in: record */
 	dict_index_t*	index,	/* in: index of record */
 	que_thr_t* 	thr)	/* in: query thread */
@ -1886,20 +1910,24 @@ lock_rec_lock_slow(
 	ulint	err;

 	ut_ad(mutex_own(&kernel_mutex));
-	ut_ad((mode == LOCK_X) || (mode == LOCK_S));
+	ut_ad((LOCK_MODE_MASK & mode) != LOCK_S
+		|| lock_table_has(thr_get_trx(thr), index->table, LOCK_IS));
+	ut_ad((LOCK_MODE_MASK & mode) != LOCK_X
+		|| lock_table_has(thr_get_trx(thr), index->table, LOCK_IX));
+	ut_ad((LOCK_MODE_MASK & mode) == LOCK_S
+		|| (LOCK_MODE_MASK & mode) == LOCK_X);
+	ut_ad(mode - (LOCK_MODE_MASK & mode) == LOCK_GAP
+			|| mode - (LOCK_MODE_MASK & mode) == 0
+			|| mode - (LOCK_MODE_MASK & mode) == LOCK_REC_NOT_GAP);
 			
 	trx = thr_get_trx(thr);
 		
-	ut_ad((mode != LOCK_S) || lock_table_has(trx, index->table,
-								LOCK_IS));
-	ut_ad((mode != LOCK_X) || lock_table_has(trx, index->table,
-								LOCK_IX));
 	if (lock_rec_has_expl(mode, rec, trx)) {
 		/* The trx already has a strong enough lock on rec: do
 		nothing */

 		err = DB_SUCCESS;
-	} else if (lock_rec_other_has_conflicting(mode, 0, 0, rec, trx)) {
+	} else if (lock_rec_other_has_conflicting(mode, rec, trx)) {

 		/* If another transaction has a non-gap conflicting request in
 		the queue, as this transaction does not have a lock strong
@ -1935,7 +1963,8 @@ lock_rec_lock(
 	ibool		impl,	/* in: if TRUE, no lock is set if no wait is
 				necessary: we assume that the caller will set
 				an implicit lock */
-	ulint		mode,	/* in: lock mode */
+	ulint		mode,	/* in: lock mode: LOCK_X or LOCK_S possibly
+				ORed to either LOCK_GAP or LOCK_REC_NOT_GAP */
 	rec_t*		rec,	/* in: record */
 	dict_index_t*	index,	/* in: index of record */
 	que_thr_t* 	thr)	/* in: query thread */
@ -1943,10 +1972,15 @@ lock_rec_lock(
 	ulint	err;

 	ut_ad(mutex_own(&kernel_mutex));
-	ut_ad((mode != LOCK_S) || lock_table_has(thr_get_trx(thr),
-						index->table, LOCK_IS));
-	ut_ad((mode != LOCK_X) || lock_table_has(thr_get_trx(thr),
-						index->table, LOCK_IX));
+	ut_ad((LOCK_MODE_MASK & mode) != LOCK_S
+		|| lock_table_has(thr_get_trx(thr), index->table, LOCK_IS));
+	ut_ad((LOCK_MODE_MASK & mode) != LOCK_X
+		|| lock_table_has(thr_get_trx(thr), index->table, LOCK_IX));
+	ut_ad((LOCK_MODE_MASK & mode) == LOCK_S
+		|| (LOCK_MODE_MASK & mode) == LOCK_X);
+	ut_ad(mode - (LOCK_MODE_MASK & mode) == LOCK_GAP
+			|| mode - (LOCK_MODE_MASK & mode) == LOCK_REC_NOT_GAP
+			|| mode - (LOCK_MODE_MASK & mode) == 0);
 			
 	if (lock_rec_lock_fast(impl, mode, rec, index, thr)) {

@ -2030,7 +2064,14 @@ lock_grant(
 					ut_dulint_get_low(lock->trx->id));
 	}

+	/* If we are resolving a deadlock by choosing another transaction
+	as a victim, then our original transaction may not be in the
+	TRX_QUE_LOCK_WAIT state, and there is no need to end the lock wait
+	for it */
+	
+	if (lock->trx->que_state == TRX_QUE_LOCK_WAIT) {	
 		trx_end_lock_wait(lock->trx);
+	}
 }

 /*****************************************************************
@ -2199,9 +2240,10 @@ lock_rec_reset_and_release_wait(
 }	

 /*****************************************************************
-Makes a record to inherit the locks of another record as gap type locks, but
-does not reset the lock bits of the other record. Also waiting lock requests
-on rec are inherited as GRANTED gap locks. */
+Makes a record to inherit the locks (except LOCK_INSERT_INTENTION type)
+of another record as gap type locks, but does not reset the lock bits of
+the other record. Also waiting lock requests on rec are inherited as
+GRANTED gap locks. */

 void
 lock_rec_inherit_to_gap(
@ -2217,9 +2259,45 @@ lock_rec_inherit_to_gap(
 	lock = lock_rec_get_first(rec);

 	while (lock != NULL) {
-		lock_rec_add_to_queue(((lock->type_mode | LOCK_GAP)
-					& ~LOCK_WAIT),
+		if (!lock_rec_get_insert_intention(lock)) {
+			
+			lock_rec_add_to_queue(LOCK_REC | lock_get_mode(lock)
+						| LOCK_GAP,
 	 			     		heir, lock->index, lock->trx);
+	 	}
+	 	
+		lock = lock_rec_get_next(rec, lock);
+	}
+}	
+
+/*****************************************************************
+Makes a record to inherit the gap locks (except LOCK_INSERT_INTENTION type)
+of another record as gap type locks, but does not reset the lock bits of the
+other record. Also waiting lock requests are inherited as GRANTED gap locks. */
+
+void
+lock_rec_inherit_to_gap_if_gap_lock(
+/*================================*/
+	rec_t*	heir,	/* in: record which inherits */
+	rec_t*	rec)	/* in: record from which inherited; does NOT reset
+			the locks on this record */
+{
+	lock_t*	lock;
+	
+	ut_ad(mutex_own(&kernel_mutex));
+	
+	lock = lock_rec_get_first(rec);
+
+	while (lock != NULL) {
+		if (!lock_rec_get_insert_intention(lock)
+		    && (page_rec_is_supremum(rec)
+			|| !lock_rec_get_rec_not_gap(lock))) {
+			
+			lock_rec_add_to_queue(LOCK_REC | lock_get_mode(lock)
+						| LOCK_GAP,
+	 			     		heir, lock->index, lock->trx);
+	 	}
+
 		lock = lock_rec_get_next(rec, lock);
 	}
 }	
@ -2778,9 +2856,10 @@ lock_update_insert(
 {
 	lock_mutex_enter_kernel();

-	/* Inherit the locks for rec, in gap mode, from the next record */
+	/* Inherit the gap-locking locks for rec, in gap mode, from the next
+	record */

-	lock_rec_inherit_to_gap(rec, page_rec_get_next(rec));
+	lock_rec_inherit_to_gap_if_gap_lock(rec, page_rec_get_next(rec));

 	lock_mutex_exit_kernel();
 }	
@ -2859,20 +2938,23 @@ static
 ibool
 lock_deadlock_occurs(
 /*=================*/
-			/* out: TRUE if a deadlock was detected */
+			/* out: TRUE if a deadlock was detected and we
+			chose trx as a victim; FALSE if no deadlock, or
+			there was a deadlock, but we chose other
+			transaction(s) as victim(s) */
 	lock_t*	lock,	/* in: lock the transaction is requesting */
 	trx_t*	trx)	/* in: transaction */
 {
 	dict_table_t*	table;
 	dict_index_t*	index;
 	trx_t*		mark_trx;
-	ibool		ret;
+	ulint		ret;
 	ulint		cost	= 0;
 	char*		err_buf;

 	ut_ad(trx && lock);
 	ut_ad(mutex_own(&kernel_mutex));
-
+retry:
 	/* We check that adding this trx to the waits-for graph
 	does not produce a cycle. First mark all active transactions
 	with 0: */
@ -2886,7 +2968,14 @@ lock_deadlock_occurs(

 	ret = lock_deadlock_recursive(trx, trx, lock, &cost);

-	if (ret) {
+	if (ret == LOCK_VICTIM_IS_OTHER) {
+		/* We chose some other trx as a victim: retry if there still
+		is a deadlock */
+
+		goto retry;
+	}
+
+	if (ret == LOCK_VICTIM_IS_START) {
 		if (lock_get_type(lock) == LOCK_TABLE) {
 			table = lock->un_member.tab_lock.table;
 			index = NULL;
@ -2899,19 +2988,6 @@ lock_deadlock_occurs(

 		err_buf = lock_latest_err_buf + strlen(lock_latest_err_buf);
 			
-		err_buf += sprintf(err_buf,
-		"*** (2) WAITING FOR THIS LOCK TO BE GRANTED:\n");
-
-		ut_a(err_buf <= lock_latest_err_buf + 4000);
-			
-		if (lock_get_type(lock) == LOCK_REC) {
-			lock_rec_print(err_buf, lock);
-			err_buf += strlen(err_buf);
-		} else {
-			lock_table_print(err_buf, lock);
-			err_buf += strlen(err_buf);
-		}
-			
 		ut_a(err_buf <= lock_latest_err_buf + 4000);

 		err_buf += sprintf(err_buf,
@ -2923,30 +2999,39 @@ lock_deadlock_occurs(
 		sess_raise_error_low(trx, DB_DEADLOCK, lock->type_mode, table,
 						index, NULL, NULL, NULL);
 		*/
+
+		return(TRUE);
 	}
 	
-	return(ret);
+	return(FALSE);
 }

 /************************************************************************
 Looks recursively for a deadlock. */
 static
-ibool
+ulint
 lock_deadlock_recursive(
 /*====================*/
-				/* out: TRUE if a deadlock was detected
-				or the calculation took too long */
+				/* out: 0 if no deadlock found,
+				LOCK_VICTIM_IS_START if there was a deadlock
+				and we chose 'start' as the victim,
+				LOCK_VICTIM_IS_OTHER if a deadlock
+				was found and we chose some other trx as a
+				victim: we must do the search again in this
+				last case because there may be another
+				deadlock! */
 	trx_t*	start,		/* in: recursion starting point */
 	trx_t*	trx,		/* in: a transaction waiting for a lock */
 	lock_t*	wait_lock,	/* in: the lock trx is waiting to be granted */
 	ulint*	cost)		/* in/out: number of calculation steps thus
 				far: if this exceeds LOCK_MAX_N_STEPS_...
-				we return TRUE */
+				we return LOCK_VICTIM_IS_START */
 {
 	lock_t*	lock;
 	ulint	bit_no;
 	trx_t*	lock_trx;
 	char*	err_buf;
+	ulint	ret;
 	
 	ut_a(trx && start && wait_lock);
 	ut_ad(mutex_own(&kernel_mutex));
@ -2955,14 +3040,14 @@ lock_deadlock_recursive(
 		/* We have already exhaustively searched the subtree starting
 		from this trx */

-		return(FALSE);
+		return(0);
 	}

 	*cost = *cost + 1;

 	if (*cost > LOCK_MAX_N_STEPS_IN_DEADLOCK_CHECK) {

-		return(TRUE);
+		return(LOCK_VICTIM_IS_START);
 	}

 	lock = wait_lock;
@ -2998,6 +3083,9 @@ lock_deadlock_recursive(
 			lock_trx = lock->trx;

 			if (lock_trx == start) {
+				/* We came back to the recursion starting
+				point: a deadlock detected */
+				
 				err_buf = lock_latest_err_buf;

 				ut_sprintf_timestamp(err_buf);
@ -3045,11 +3133,60 @@ lock_deadlock_recursive(
 			
 				ut_a(err_buf <= lock_latest_err_buf + 4000);

+				err_buf += sprintf(err_buf,
+			"*** (2) WAITING FOR THIS LOCK TO BE GRANTED:\n");
+
+				ut_a(err_buf <= lock_latest_err_buf + 4000);
+			
+				if (lock_get_type(start->wait_lock)
+								== LOCK_REC) {
+					lock_rec_print(err_buf,
+							start->wait_lock);
+					err_buf += strlen(err_buf);
+				} else {
+					lock_table_print(err_buf,
+							start->wait_lock);
+					err_buf += strlen(err_buf);
+				}
+
 				if (lock_print_waits) {
 					printf("Deadlock detected\n");
 				}

-				return(TRUE);
+				if (ut_dulint_cmp(wait_lock->trx->undo_no,
+							start->undo_no) >= 0) {
+					/* Our recursion starting point
+					transaction is 'smaller', let us
+					choose 'start' as the victim and roll
+					back it */
+
+					return(LOCK_VICTIM_IS_START);
+				}		
+
+				lock_deadlock_found = TRUE;
+
+				ut_a(err_buf <= lock_latest_err_buf + 4000);
+
+				/* Let us choose the transaction of wait_lock
+				as a victim to try to avoid deadlocking our
+				recursion starting point transaction */
+				
+				err_buf += sprintf(err_buf,
+				"*** WE ROLL BACK TRANSACTION (1)\n");
+				
+				wait_lock->trx->was_chosen_as_deadlock_victim
+								= TRUE;
+				
+				lock_cancel_waiting_and_release(wait_lock);
+
+				/* Since trx and wait_lock are no longer
+				in the waits-for graph, we can return FALSE;
+				note that our selective algorithm can choose
+				several transactions as victims, but still
+				we may end up rolling back also the recursion
+				starting point transaction! */
+
+				return(LOCK_VICTIM_IS_OTHER);
 			}
 	
 			if (lock_trx->que_state == TRX_QUE_LOCK_WAIT) {
@ -3058,10 +3195,11 @@ lock_deadlock_recursive(
 				incompatible mode, and is itself waiting for
 				a lock */

-				if (lock_deadlock_recursive(start, lock_trx,
-						lock_trx->wait_lock, cost)) {
+				ret = lock_deadlock_recursive(start, lock_trx,
+						lock_trx->wait_lock, cost);
+				if (ret != 0) {

-					return(TRUE);
+					return(ret);
 				}
 			}
 		}
@ -3153,12 +3291,16 @@ lock_table_remove_low(
 /*************************************************************************
 Enqueues a waiting request for a table lock which cannot be granted
 immediately. Checks for deadlocks. */
-
+static
 ulint
 lock_table_enqueue_waiting(
 /*=======================*/
 				/* out: DB_LOCK_WAIT, DB_DEADLOCK, or
-				DB_QUE_THR_SUSPENDED */
+				DB_QUE_THR_SUSPENDED, or DB_SUCCESS;
+				DB_SUCCESS means that there was a deadlock,
+				but another transaction was chosen as a
+				victim, and we got the lock immediately:
+				no need to wait then */
 	ulint		mode,	/* in: lock mode this transaction is
 				requesting */
 	dict_table_t*	table,	/* in: table */
@ -3205,7 +3347,15 @@ table->name);
 		return(DB_DEADLOCK);
 	}

+	if (trx->wait_lock == NULL) {
+		/* Deadlock resolution chose another transaction as a victim,
+		and we accidentally got our lock granted! */
+	
+		return(DB_SUCCESS);
+	}
+	
 	trx->que_state = TRX_QUE_LOCK_WAIT;
+	trx->was_chosen_as_deadlock_victim = FALSE;
 	trx->wait_started = time(NULL);

 	ut_a(que_thr_stop(thr));
@ -3292,7 +3442,7 @@ lock_table(
 	if (lock_table_other_has_incompatible(trx, LOCK_WAIT, table, mode)) {
 	
 		/* Another trx has a request on the table in an incompatible
-		mode: this trx must wait */
+		mode: this trx may have to wait */

 		err = lock_table_enqueue_waiting(mode, table, thr);
 			
@ -3659,7 +3809,11 @@ lock_rec_print(
 	}

 	if (lock_rec_get_gap(lock)) {
-		buf += sprintf(buf, " gap type lock");
+		buf += sprintf(buf, " locks gap before rec");
+	}
+
+	if (lock_rec_get_rec_not_gap(lock)) {
+		buf += sprintf(buf, " locks rec but not gap");
 	}

 	if (lock_rec_get_insert_intention(lock)) {
@ -4080,7 +4234,8 @@ lock_rec_queue_validate(
 		if (impl_trx && lock_rec_other_has_expl_req(LOCK_S, 0,
 				LOCK_WAIT, rec, impl_trx)) {

-			ut_a(lock_rec_has_expl(LOCK_X, rec, impl_trx));
+			ut_a(lock_rec_has_expl(LOCK_X | LOCK_REC_NOT_GAP, rec,
+								impl_trx));
 		}
 	}

@ -4095,7 +4250,8 @@ lock_rec_queue_validate(
 		if (impl_trx && lock_rec_other_has_expl_req(LOCK_S, 0,
 				LOCK_WAIT, rec, impl_trx)) {

-			ut_a(lock_rec_has_expl(LOCK_X, rec, impl_trx));
+			ut_a(lock_rec_has_expl(LOCK_X | LOCK_REC_NOT_GAP, rec,
+								impl_trx));
 		}
 	}

@ -4359,8 +4515,8 @@ lock_rec_insert_check_and_lock(

 	*inherit = TRUE;

-	/* If another transaction has an explicit lock request, gap or not,
-	waiting or granted, on the successor, the insert has to wait.
+	/* If another transaction has an explicit lock request which locks
+	the gap, waiting or granted, on the successor, the insert has to wait.

 	An exception is the case where the lock by the another transaction
 	is a gap type lock which it placed to wait for its turn to insert. We
@ -4369,8 +4525,10 @@ lock_rec_insert_check_and_lock(
 	had to wait for their insert. Both had waiting gap type lock requests
 	on the successor, which produced an unnecessary deadlock. */

-	if (lock_rec_other_has_conflicting(LOCK_X, LOCK_GAP,
-				LOCK_INSERT_INTENTION, next_rec, trx)) {
+	if (lock_rec_other_has_conflicting(LOCK_X | LOCK_GAP
+				| LOCK_INSERT_INTENTION, next_rec, trx)) {
+
+		/* Note that we may get DB_SUCCESS also here! */
 		err = lock_rec_enqueue_waiting(LOCK_X | LOCK_GAP
 						| LOCK_INSERT_INTENTION,
 						next_rec, index, thr);
@ -4418,9 +4576,11 @@ lock_rec_convert_impl_to_expl(
 		/* If the transaction has no explicit x-lock set on the
 		record, set one for it */

-		if (!lock_rec_has_expl(LOCK_X, rec, impl_trx)) {
+		if (!lock_rec_has_expl(LOCK_X | LOCK_REC_NOT_GAP, rec,
+								impl_trx)) {

-			lock_rec_add_to_queue(LOCK_REC | LOCK_X, rec, index,
+			lock_rec_add_to_queue(LOCK_REC | LOCK_X
+					      | LOCK_REC_NOT_GAP, rec, index,
 								impl_trx);
 		}
 	}
@ -4466,7 +4626,7 @@ lock_clust_rec_modify_check_and_lock(

 	lock_rec_convert_impl_to_expl(rec, index);

-	err = lock_rec_lock(TRUE, LOCK_X, rec, index, thr);
+	err = lock_rec_lock(TRUE, LOCK_X | LOCK_REC_NOT_GAP, rec, index, thr);

 	lock_mutex_exit_kernel();

@ -4511,7 +4671,7 @@ lock_sec_rec_modify_check_and_lock(

 	ut_ad(lock_table_has(thr_get_trx(thr), index->table, LOCK_IX));

-	err = lock_rec_lock(TRUE, LOCK_X, rec, index, thr);
+	err = lock_rec_lock(TRUE, LOCK_X | LOCK_REC_NOT_GAP, rec, index, thr);

 	lock_mutex_exit_kernel();
 	
@ -4545,6 +4705,8 @@ lock_sec_rec_read_check_and_lock(
 	ulint		mode,	/* in: mode of the lock which the read cursor
 				should set on records: LOCK_S or LOCK_X; the
 				latter is possible in SELECT FOR UPDATE */
+	ulint		gap_mode,/* in: LOCK_ORDINARY, LOCK_GAP, or
+				LOCK_REC_NOT_GAP */
 	que_thr_t*	thr)	/* in: query thread */
 {
 	ulint	err;
@ -4576,7 +4738,7 @@ lock_sec_rec_read_check_and_lock(
 		lock_rec_convert_impl_to_expl(rec, index);
 	}

-	err = lock_rec_lock(FALSE, mode, rec, index, thr);
+	err = lock_rec_lock(FALSE, mode | gap_mode, rec, index, thr);

 	lock_mutex_exit_kernel();

@ -4607,13 +4769,16 @@ lock_clust_rec_read_check_and_lock(
 	ulint		mode,	/* in: mode of the lock which the read cursor
 				should set on records: LOCK_S or LOCK_X; the
 				latter is possible in SELECT FOR UPDATE */
+	ulint		gap_mode,/* in: LOCK_ORDINARY, LOCK_GAP, or
+				LOCK_REC_NOT_GAP */
 	que_thr_t*	thr)	/* in: query thread */
 {
 	ulint	err;

 	ut_ad(index->type & DICT_CLUSTERED);
 	ut_ad(page_rec_is_user_rec(rec) || page_rec_is_supremum(rec));
-	
+	ut_ad(gap_mode == LOCK_ORDINARY || gap_mode == LOCK_GAP
+					|| gap_mode == LOCK_REC_NOT_GAP);
 	if (flags & BTR_NO_LOCKING_FLAG) {

 		return(DB_SUCCESS);
@ -4631,7 +4796,7 @@ lock_clust_rec_read_check_and_lock(
 		lock_rec_convert_impl_to_expl(rec, index);
 	}

-	err = lock_rec_lock(FALSE, mode, rec, index, thr);
+	err = lock_rec_lock(FALSE, mode | gap_mode, rec, index, thr);

 	lock_mutex_exit_kernel();

--- a/innobase/log/log0log.c
+++ b/innobase/log/log0log.c
@ -1654,8 +1654,8 @@ log_reset_first_header_and_checkpoint(
 	lsn = ut_dulint_add(start, LOG_BLOCK_HDR_SIZE);

 	/* Write the label of ibbackup --restore */
-	sprintf(hdr_buf + LOG_FILE_WAS_CREATED_BY_HOT_BACKUP, "ibbackup ");
-	ut_sprintf_timestamp(hdr_buf + LOG_FILE_WAS_CREATED_BY_HOT_BACKUP
+	sprintf((char*) hdr_buf + LOG_FILE_WAS_CREATED_BY_HOT_BACKUP, "ibbackup ");
+	ut_sprintf_timestamp((char*) hdr_buf + LOG_FILE_WAS_CREATED_BY_HOT_BACKUP
 						+ strlen("ibbackup "));
 	buf = hdr_buf + LOG_CHECKPOINT_1;
 	
--- a/innobase/log/log0recv.c
+++ b/innobase/log/log0recv.c
@ -69,6 +69,8 @@ ulint	recv_previous_parsed_rec_type	= 999999;
 ulint	recv_previous_parsed_rec_offset	= 0;
 ulint	recv_previous_parsed_rec_is_multi = 0;

+ulint	recv_max_parsed_page_no		= 0;
+
 /************************************************************
 Creates the recovery system. */

@ -141,7 +143,13 @@ recv_sys_empty_hash(void)
 /*=====================*/
 {
 	ut_ad(mutex_own(&(recv_sys->mutex)));
-	ut_a(recv_sys->n_addrs == 0);
+	if (recv_sys->n_addrs != 0) {
+		fprintf(stderr,
+"InnoDB: Error: %lu pages with log records were left unprocessed!\n"
+"InnoDB: Maximum page number with log records on it %lu\n",
+			recv_sys->n_addrs, recv_max_parsed_page_no);
+		ut_a(0);
+	}
 	
 	hash_table_free(recv_sys->addr_hash);
 	mem_heap_empty(recv_sys->heap);
@ -1361,6 +1369,14 @@ recv_apply_log_recs_for_backup(
 		n_pages_total += file_sizes[i];
 	}

+	if (recv_max_parsed_page_no >= n_pages_total) {
+		printf(
+"InnoDB: Error: tablespace size %lu pages, but a log record on page %lu!\n"
+"InnoDB: Are you sure you have specified all the ibdata files right in\n"
+"InnoDB: the my.cnf file you gave as the argument to ibbackup --restore?\n",
+			n_pages_total, recv_max_parsed_page_no);
+	}
+
 	printf( 
 "InnoDB: Starting an apply batch of log records to the database...\n"
 "InnoDB: Progress in percents: ");
@ -1381,7 +1397,7 @@ recv_apply_log_recs_for_backup(
 							&success);
 			if (!success) {
 				printf(
-"InnoDB: Error: cannot open %lu'th data file %s\n", nth_file);
+"InnoDB: Error: cannot open %lu'th data file\n", nth_file);

 				exit(1);
 			}
@ -1397,7 +1413,7 @@ recv_apply_log_recs_for_backup(
 				UNIV_PAGE_SIZE);
 			if (!success) {
 				printf(
-"InnoDB: Error: cannot read page no %lu from %lu'th data file %s\n",
+"InnoDB: Error: cannot read page no %lu from %lu'th data file\n",
 				nth_page_in_file, nth_file);

 				exit(1);
@ -1425,7 +1441,7 @@ recv_apply_log_recs_for_backup(
 				UNIV_PAGE_SIZE);
 			if (!success) {
 				printf(
-"InnoDB: Error: cannot write page no %lu to %lu'th data file %s\n",
+"InnoDB: Error: cannot write page no %lu to %lu'th data file\n",
 				nth_page_in_file, nth_file);

 				exit(1);
@ -1701,6 +1717,10 @@ recv_parse_log_rec(
 		return(0);
 	}

+	if (*page_no > recv_max_parsed_page_no) {
+		recv_max_parsed_page_no = *page_no;
+	}
+	
 	return(new_ptr - ptr);
 }

@ -1779,7 +1799,7 @@ recv_report_corrupt_log(
 "InnoDB: Recv offset %lu, prev %lu\n",
 		recv_previous_parsed_rec_type,
 		recv_previous_parsed_rec_is_multi,
-		ptr - recv_sys->buf,
+		(ulint)(ptr - recv_sys->buf),
 		recv_previous_parsed_rec_offset);

 	if ((ulint)(ptr - recv_sys->buf + 100)
--- a/innobase/mem/mem0dbg.c
+++ b/innobase/mem/mem0dbg.c
@ -350,6 +350,16 @@ mem_hash_remove(
 							node->nth_heap);
 	   printf("in %s line %lu and tried to free in %s line %lu.\n",
 	  			node->file_name, node->line, file_name, line);
+
+	   printf(
+	   "Hex dump of 400 bytes around memory heap first block start:\n");
+
+	   ut_print_buf((byte*)(node->heap) - 200, 400);
+
+	   printf("\nDump of the mem heap:\n");
+
+	   mem_heap_validate_or_print(node->heap, NULL, TRUE, &error, &size,
+								NULL, NULL);
 	   ut_error;
 	}

--- a/innobase/os/os0file.c
+++ b/innobase/os/os0file.c
@ -148,7 +148,7 @@ Gets the operating system version. Currently works only on Windows. */
 ulint
 os_get_os_version(void)
 /*===================*/
-                  /* out: OS_WIN95, OS_WIN31, OS_WINNT (2000 == NT) */
+                  /* out: OS_WIN95, OS_WIN31, OS_WINNT, OS_WIN2000 */
 {
 #ifdef __WIN__
  	OSVERSIONINFO     os_info;
@ -162,7 +162,11 @@ os_get_os_version(void)
  	} else if (os_info.dwPlatformId == VER_PLATFORM_WIN32_WINDOWS) {
    		return(OS_WIN95);
  	} else if (os_info.dwPlatformId == VER_PLATFORM_WIN32_NT) {
+		if (os_info.dwMajorVersion <= 4) {
    			return(OS_WINNT);
+    		} else {
+			return(OS_WIN2000);
+    		}
  	} else {
    		ut_error;
    		return(0);
@ -268,9 +272,7 @@ os_file_get_last_error(void)
 }

 /********************************************************************
-Does error handling when a file operation fails. If we have run out
-of disk space, then the user can clean the disk. If we do not find
-a specified file, then the user can copy it to disk. */
+Does error handling when a file operation fails. */
 static
 ibool
 os_file_handle_error(
@ -503,7 +505,11 @@ try_again:
 		        value 2 denotes that we do not flush the log at every
 		        commit, but only once per second */
 		} else {
-		        attributes = attributes | FILE_FLAG_NO_BUFFERING;
+			if (srv_win_file_flush_method ==
+					SRV_WIN_IO_UNBUFFERED) {
+		        	attributes = attributes
+						| FILE_FLAG_NO_BUFFERING;
+			}
 		}
 #endif
 	} else if (purpose == OS_FILE_NORMAL) {
@ -514,7 +520,11 @@ try_again:
 		        value 2 denotes that we do not flush the log at every
 		        commit, but only once per second */
 		} else {
-		        attributes = attributes | FILE_FLAG_NO_BUFFERING;
+			if (srv_win_file_flush_method ==
+					SRV_WIN_IO_UNBUFFERED) {
+		        	attributes = attributes
+						| FILE_FLAG_NO_BUFFERING;
+			}
 		}
 #endif
 	} else {
@ -1752,6 +1762,7 @@ os_aio(
 	os_aio_array_t*	array;
 	os_aio_slot_t*	slot;
 #ifdef WIN_ASYNC_IO
+	ibool		retval;
 	BOOL		ret		= TRUE;
 	DWORD		len		= n;
 	void*		dummy_mess1;
@ -1824,6 +1835,8 @@ try_again:
 		if (os_aio_use_native_aio) {
 #ifdef WIN_ASYNC_IO
 			os_n_file_reads++;
+			os_bytes_read_since_printout += len;
+			
 			ret = ReadFile(file, buf, (DWORD)n, &len,
 							&(slot->control));
 #elif defined(POSIX_ASYNC_IO)
@ -1870,10 +1883,12 @@ try_again:
 	    		    where we also use async i/o: in Windows we must
 	    		    use the same wait mechanism as for async i/o */
 	    		
-	    		    return(os_aio_windows_handle(ULINT_UNDEFINED,
+	    		    retval = os_aio_windows_handle(ULINT_UNDEFINED,
 					slot->pos,
 		    			&dummy_mess1, &dummy_mess2,
-					&dummy_type));
+					&dummy_type);
+
+			    return(retval);
 	    		}

 			return(TRUE);
@ -1897,8 +1912,6 @@ try_again:
 		goto try_again;
 	}	

-	ut_error;
-	
 	return(FALSE);
 }

@ -1958,14 +1971,14 @@ os_aio_windows_handle(
 	n = array->n_slots / array->n_segments;

 	if (array == os_aio_sync_array) {
-		srv_io_thread_op_info[orig_seg] = "wait windows aio for 1 page";
+		srv_io_thread_op_info[orig_seg] = "wait Windows aio for 1 page";

 		ut_ad(pos < array->n_slots); 
 		os_event_wait(array->events[pos]);
 		i = pos;
 	} else {
 		srv_io_thread_op_info[orig_seg] =
-						"wait windows aio for n pages";
+						"wait Windows aio";
 		i = os_event_wait_multiple(n, (array->events) + segment * n);
 	}

@ -1991,9 +2004,7 @@ os_aio_windows_handle(
 		         ut_a(TRUE == os_file_flush(slot->file));
 		}
 	} else {
-		os_file_get_last_error();
-
-		ut_error;
+		os_file_handle_error(slot->file, slot->name);
 		
 		ret_val = FALSE;
 	}		  
--- a/innobase/os/os0proc.c
+++ b/innobase/os/os0proc.c
@ -18,6 +18,23 @@ Created 9/30/1995 Heikki Tuuri

 #include "ut0mem.h"

+/********************************************************************
+Converts the current process id to a number. It is not guaranteed that the
+number is unique. In Linux returns the 'process number' of the current
+thread. That number is the same as one sees in 'top', for example. In Linux
+the thread id is not the same as one sees in 'top'. */
+
+ulint
+os_proc_get_number(void)
+/*====================*/
+{
+#ifdef __WIN__
+	return((ulint)GetCurrentProcessId());
+#else
+	return((ulint)getpid());
+#endif
+}
+
 /********************************************************************
 Allocates non-cacheable memory. */

--- a/innobase/os/os0sync.c
+++ b/innobase/os/os0sync.c
@ -66,8 +66,12 @@ os_event_create(
 	event = ut_malloc(sizeof(struct os_event_struct));

 	os_fast_mutex_init(&(event->os_mutex));
-	pthread_cond_init(&(event->cond_var), NULL);

+#if defined(UNIV_HOTBACKUP) && defined(UNIV_HPUX10)
+	pthread_cond_init(&(event->cond_var), pthread_condattr_default);
+#else
+	pthread_cond_init(&(event->cond_var), NULL);
+#endif
 	event->is_set = FALSE;

 	return(event);
@ -440,9 +444,13 @@ os_fast_mutex_init(
 	ut_a(fast_mutex);
 	
 	InitializeCriticalSection((LPCRITICAL_SECTION) fast_mutex);
+#else
+#if defined(UNIV_HOTBACKUP) && defined(UNIV_HPUX10)
+	pthread_mutex_init(fast_mutex, pthread_mutexattr_default);
 #else
 	pthread_mutex_init(fast_mutex, MY_MUTEX_INIT_FAST);
 #endif
+#endif
 }

 /**************************************************************
--- a/innobase/os/os0thread.c
+++ b/innobase/os/os0thread.c
@ -126,7 +126,9 @@ os_thread_create(
 	os_thread_t	pthread;
 	pthread_attr_t  attr;

+#if !(defined(UNIV_HOTBACKUP) && defined(UNIV_HPUX10))
        pthread_attr_init(&attr);
+#endif
        
 #ifdef UNIV_AIX
 	/* We must make sure a thread stack is at least 32 kB, otherwise
@ -142,16 +144,21 @@ os_thread_create(
 		 exit(1);
 	}
 #endif
-	ret = pthread_create(&pthread, &attr, start_f, arg);

+#if defined(UNIV_HOTBACKUP) && defined(UNIV_HPUX10)
+	ret = pthread_create(&pthread, pthread_attr_default, start_f, arg);
+#else
+	ret = pthread_create(&pthread, &attr, start_f, arg);
+#endif
        if (ret) {
 	         fprintf(stderr,
          "InnoDB: Error: pthread_create returned %d\n", ret);
 		 exit(1);
 	}

+#if !(defined(UNIV_HOTBACKUP) && defined(UNIV_HPUX10))
 	pthread_attr_destroy(&attr);
-
+#endif
 	if (srv_set_thread_priorities) {
 	
 	        my_pthread_setprio(pthread, srv_query_thread_priority);
--- a/innobase/page/page0cur.c
+++ b/innobase/page/page0cur.c
@ -169,7 +169,7 @@ page_cur_search_with_match(
 	ut_ad(dtuple_check_typed(tuple));
 	ut_ad((mode == PAGE_CUR_L) || (mode == PAGE_CUR_LE)
 	      || (mode == PAGE_CUR_G) || (mode == PAGE_CUR_GE)
-	      || (mode == PAGE_CUR_DBG));
+	      || (mode == PAGE_CUR_LE_OR_EXTENDS) || (mode == PAGE_CUR_DBG));
 	      
 #ifdef PAGE_CUR_ADAPT
 	if ((page_header_get_field(page, PAGE_LEVEL) == 0)
@ -232,9 +232,26 @@ page_cur_search_with_match(
 			low_matched_bytes = cur_matched_bytes;

 		} else if (cmp == -1) {
+
+			if (mode == PAGE_CUR_LE_OR_EXTENDS
+			    && dfield_get_len(dtuple_get_nth_field(tuple,
+			    				cur_matched_fields))
+			    	== cur_matched_bytes
+			    && rec_get_nth_field_len(mid_rec,
+							cur_matched_fields)
+				!= UNIV_SQL_NULL) {
+
+				/* This means current dfield is not SQL
+			    	NULL, and the current rec field extends it */
+
+				low = mid;
+				low_matched_fields = cur_matched_fields;
+				low_matched_bytes = cur_matched_bytes;
+			} else {
 				up = mid;
 				up_matched_fields = cur_matched_fields;
 				up_matched_bytes = cur_matched_bytes;
+			}

 		} else if ((mode == PAGE_CUR_G) || (mode == PAGE_CUR_LE)) {
 			low = mid;
@ -252,8 +269,8 @@ page_cur_search_with_match(
 	slot = page_dir_get_nth_slot(page, up);
 	up_rec = page_dir_slot_get_rec(slot);

-	/* Perform linear search until the upper and lower records
-	come to distance 1 of each other. */
+	/* Perform linear search until the upper and lower records come to
+	distance 1 of each other. */

   	while (page_rec_get_next(low_rec) != up_rec) {

@ -272,10 +289,25 @@ page_cur_search_with_match(
 			low_matched_bytes = cur_matched_bytes;

 		} else if (cmp == -1) {
+			if (mode == PAGE_CUR_LE_OR_EXTENDS
+			    && dfield_get_len(dtuple_get_nth_field(tuple,
+			    				cur_matched_fields))
+			    	== cur_matched_bytes
+			    && rec_get_nth_field_len(mid_rec,
+							cur_matched_fields)
+				!= UNIV_SQL_NULL) {
+
+				/* This means current dfield is not SQL
+			    	NULL, and the current rec field extends it */
+
+				low = mid;
+				low_matched_fields = cur_matched_fields;
+				low_matched_bytes = cur_matched_bytes;
+			} else {
 				up_rec = mid_rec;
 				up_matched_fields = cur_matched_fields;
 				up_matched_bytes = cur_matched_bytes;
-
+			}
 		} else if ((mode == PAGE_CUR_G) || (mode == PAGE_CUR_LE)) {
 			low_rec = mid_rec;
 			low_matched_fields = cur_matched_fields;
--- a/innobase/page/page0page.c
+++ b/innobase/page/page0page.c
@ -1299,12 +1299,16 @@ page_rec_validate(
 	heap_no = rec_get_heap_no(rec);

 	if (!(n_owned <= PAGE_DIR_SLOT_MAX_N_OWNED)) {
-		fprintf(stderr, "Dir slot n owned too big %lu\n", n_owned);
+		fprintf(stderr,
+			"InnoDB: Dir slot of rec %lu, n owned too big %lu\n",
+				(ulint)(rec - page), n_owned);
 		return(FALSE);
 	}

 	if (!(heap_no < page_header_get_field(page, PAGE_N_HEAP))) {
-		fprintf(stderr, "Heap no too big %lu %lu\n", heap_no,
+		fprintf(stderr,
+		"InnoDB: Heap no of rec %lu too big %lu %lu\n",
+				(ulint)(rec - page), heap_no,
 				page_header_get_field(page, PAGE_N_HEAP));
 		return(FALSE);
 	}
@ -1312,6 +1316,194 @@ page_rec_validate(
 	return(TRUE);
 }
 	
+/*******************************************************************
+This function checks the consistency of an index page when we do not
+know the index. This is also resilient so that this should never crash
+even if the page is total garbage. */
+
+ibool
+page_simple_validate(
+/*=================*/
+			/* out: TRUE if ok */
+	page_t*	page)	/* in: index page */
+{
+	page_cur_t 	cur;
+	page_dir_slot_t* slot;
+	ulint		slot_no;
+	ulint		n_slots;
+	rec_t*		rec;
+	byte*		rec_heap_top;
+	ulint		count;
+	ulint		own_count;
+	ibool		ret	= FALSE;
+
+	/* Check first that the record heap and the directory do not
+	overlap. */
+
+	n_slots = page_dir_get_n_slots(page);
+
+	if (n_slots > UNIV_PAGE_SIZE / 4) {
+		fprintf(stderr,
+	"InnoDB: Nonsensical number %lu of page dir slots\n", n_slots);
+
+		goto func_exit;
+	}
+
+	rec_heap_top = page_header_get_ptr(page, PAGE_HEAP_TOP);
+	
+	if (rec_heap_top > page_dir_get_nth_slot(page, n_slots - 1)) {
+
+		fprintf(stderr,
+    "InnoDB: Record heap and dir overlap on a page, heap top %lu, dir %lu\n",
+       		(ulint)(page_header_get_ptr(page, PAGE_HEAP_TOP) - page),
+       		(ulint)(page_dir_get_nth_slot(page, n_slots - 1) - page));
+
+       		goto func_exit;
+       	}
+
+	/* Validate the record list in a loop checking also that it is
+	consistent with the page record directory. */
+
+	count = 0;
+	own_count = 1;
+	slot_no = 0;
+	slot = page_dir_get_nth_slot(page, slot_no);
+
+	page_cur_set_before_first(page, &cur);
+
+	for (;;) {
+		rec = (&cur)->rec;
+		
+		if (rec > rec_heap_top) {
+			fprintf(stderr,
+			"InnoDB: Record %lu is above rec heap top %lu\n",
+			(ulint)(rec - page), (ulint)(rec_heap_top - page));
+
+			goto func_exit;
+		}
+
+		if (rec_get_n_owned(rec) != 0) {
+			/* This is a record pointed to by a dir slot */
+			if (rec_get_n_owned(rec) != own_count) {
+
+				fprintf(stderr,
+		"InnoDB: Wrong owned count %lu, %lu, rec %lu\n",
+				rec_get_n_owned(rec), own_count,
+				(ulint)(rec - page));
+
+				goto func_exit;
+			}
+
+			if (page_dir_slot_get_rec(slot) != rec) {
+				fprintf(stderr,
+		"InnoDB: Dir slot does not point to right rec %lu\n",
+					(ulint)(rec - page));
+
+				goto func_exit;
+			}
+						
+			own_count = 0;
+
+			if (!page_cur_is_after_last(&cur)) {
+				slot_no++;
+				slot = page_dir_get_nth_slot(page, slot_no);
+			}
+		}
+
+		if (page_cur_is_after_last(&cur)) {
+
+			break;
+		}
+
+		if (rec_get_next_offs(rec) < FIL_PAGE_DATA
+				|| rec_get_next_offs(rec) >= UNIV_PAGE_SIZE) {
+			fprintf(stderr,
+		"InnoDB: Next record offset nonsensical %lu for rec %lu\n",
+			  rec_get_next_offs(rec),
+			  (ulint)(rec - page));
+
+			goto func_exit;
+		}
+
+		count++;		
+
+		if (count > UNIV_PAGE_SIZE) {
+			fprintf(stderr,
+		"InnoDB: Page record list appears to be circular %lu\n",
+								count);
+			goto func_exit;
+		}
+		
+		page_cur_move_to_next(&cur);
+		own_count++;
+	}
+	
+	if (rec_get_n_owned(rec) == 0) {
+		fprintf(stderr, "InnoDB: n owned is zero in a supremum rec\n");
+
+		goto func_exit;
+	}
+		
+	if (slot_no != n_slots - 1) {
+		fprintf(stderr, "InnoDB: n slots wrong %lu, %lu\n",
+			slot_no, n_slots - 1);
+		goto func_exit;
+	}		
+
+	if (page_header_get_field(page, PAGE_N_RECS) + 2 != count + 1) {
+		fprintf(stderr, "InnoDB: n recs wrong %lu %lu\n",
+		page_header_get_field(page, PAGE_N_RECS) + 2,  count + 1);
+
+		goto func_exit;
+	}
+
+	/* Check then the free list */
+	rec = page_header_get_ptr(page, PAGE_FREE);
+
+	while (rec != NULL) {
+		if (rec < page + FIL_PAGE_DATA
+				|| rec >= page + UNIV_PAGE_SIZE) {
+			fprintf(stderr,
+		"InnoDB: Free list record has a nonsensical offset %lu\n",
+			(ulint)(rec - page));
+
+			goto func_exit;
+		}
+
+		if (rec > rec_heap_top) {
+			fprintf(stderr,
+		"InnoDB: Free list record %lu is above rec heap top %lu\n",
+			(ulint)(rec - page), (ulint)(rec_heap_top - page));
+
+			goto func_exit;
+		}
+
+		count++;
+		
+		if (count > UNIV_PAGE_SIZE) {
+			fprintf(stderr,
+		"InnoDB: Page free list appears to be circular %lu\n",
+								count);
+			goto func_exit;
+		}
+
+		rec = page_rec_get_next(rec);
+	}
+	
+	if (page_header_get_field(page, PAGE_N_HEAP) != count + 1) {
+
+		fprintf(stderr, "InnoDB: N heap is wrong %lu, %lu\n",
+		page_header_get_field(page, PAGE_N_HEAP), count + 1);
+
+		goto func_exit;
+	}
+
+	ret = TRUE;	
+
+func_exit:
+	return(ret);			  
+}
+
 /*******************************************************************
 This function checks the consistency of an index page. */

@ -1339,6 +1531,17 @@ page_validate(
 	ulint		i;
 	char           	err_buf[1000];
 	
+	if (!page_simple_validate(page)) {
+		fprintf(stderr,
+"InnoDB: Apparent corruption in page %lu in index %s in table %s\n",
+			buf_frame_get_page_no(page), index->name,
+			index->table_name);
+
+		buf_page_print(page);
+
+		return(FALSE);
+	}
+
 	heap = mem_heap_create(UNIV_PAGE_SIZE);
 	
 	/* The following buffer is used to check that the
@ -1357,7 +1560,7 @@ page_validate(
 	if (!(page_header_get_ptr(page, PAGE_HEAP_TOP) <=
 			page_dir_get_nth_slot(page, n_slots - 1))) {
 		fprintf(stderr,
-       	"Record heap and dir overlap on a page in index %s, %lu, %lu\n",
+"InnoDB: Record heap and dir overlap on a page in index %s, %lu, %lu\n",
       		index->name, (ulint)page_header_get_ptr(page, PAGE_HEAP_TOP),
       		(ulint)page_dir_get_nth_slot(page, n_slots - 1));

@ -1385,10 +1588,14 @@ page_validate(
 		if ((count >= 2) && (!page_cur_is_after_last(&cur))) {
 			if (!(1 == cmp_rec_rec(rec, old_rec, index))) {
 				fprintf(stderr,
-					"Records in wrong order in index %s\n",
-					index->name);
+"InnoDB: Records in wrong order on page %lu index %s table %s\n",
+					buf_frame_get_page_no(page),
+					index->name,
+					index->table_name);
+
 	 		 	rec_sprintf(err_buf, 900, old_rec);
-	  			fprintf(stderr, "InnoDB: record %s\n", err_buf);
+	  			fprintf(stderr,
+				"InnoDB: previous record %s\n", err_buf);
 				
 	 		 	rec_sprintf(err_buf, 900, rec);
 	  			fprintf(stderr, "InnoDB: record %s\n", err_buf);
@ -1410,7 +1617,7 @@ page_validate(
 				/* No other record may overlap this */

 				fprintf(stderr,
-				"Record overlaps another in index %s \n",
+			"InnoDB: Record overlaps another in index %s \n",
 				index->name);

 				goto func_exit;
@ -1423,7 +1630,7 @@ page_validate(
 			/* This is a record pointed to by a dir slot */
 			if (rec_get_n_owned(rec) != own_count) {
 				fprintf(stderr,
-				"Wrong owned count %lu, %lu, in index %s\n",
+			"InnoDB: Wrong owned count %lu, %lu, in index %s\n",
 				rec_get_n_owned(rec), own_count,
 				index->name);

@ -1432,7 +1639,7 @@ page_validate(

 			if (page_dir_slot_get_rec(slot) != rec) {
 				fprintf(stderr,
-				"Dir slot does not point to right rec in %s\n",
+			"InnoDB: Dir slot does not point to right rec in %s\n",
 				index->name);

 				goto func_exit;
@ -1454,7 +1661,7 @@ page_validate(
 		if (rec_get_next_offs(rec) < FIL_PAGE_DATA
 				|| rec_get_next_offs(rec) >= UNIV_PAGE_SIZE) {
 			fprintf(stderr,
-			  "Next record offset wrong %lu in index %s\n",
+		"InnoDB: Next record offset wrong %lu in index %s\n",
 			  rec_get_next_offs(rec), index->name);

 			goto func_exit;
@ -1467,19 +1674,20 @@ page_validate(
 	}
 	
 	if (rec_get_n_owned(rec) == 0) {
-		fprintf(stderr, "n owned is zero in index %s\n", index->name);
+		fprintf(stderr,
+			"InnoDB: n owned is zero in index %s\n", index->name);

 		goto func_exit;
 	}
 		
 	if (slot_no != n_slots - 1) {
-		fprintf(stderr, "n slots wrong %lu %lu in index %s\n",
+		fprintf(stderr, "InnoDB: n slots wrong %lu %lu in index %s\n",
 			slot_no, n_slots - 1, index->name);
 		goto func_exit;
 	}		

 	if (page_header_get_field(page, PAGE_N_RECS) + 2 != count + 1) {
-		fprintf(stderr, "n recs wrong %lu %lu in index %s\n",
+		fprintf(stderr, "InnoDB: n recs wrong %lu %lu in index %s\n",
 		page_header_get_field(page, PAGE_N_RECS) + 2,  count + 1,
 		index->name);

@ -1487,7 +1695,8 @@ page_validate(
 	}

 	if (data_size != page_get_data_size(page)) {
-		fprintf(stderr, "Summed data size %lu, returned by func %lu\n",
+		fprintf(stderr,
+		"InnoDB: Summed data size %lu, returned by func %lu\n",
 			data_size, page_get_data_size(page));
 		goto func_exit;
 	}
@ -1508,7 +1717,7 @@ page_validate(

 			if (buf[offs + i] != 0) {
 				fprintf(stderr,
-	                "Record overlaps another in free list, index %s \n",
+	     	"InnoDB: Record overlaps another in free list, index %s \n",
 				index->name);

 				goto func_exit;
@ -1522,9 +1731,11 @@ page_validate(
 	
 	if (page_header_get_field(page, PAGE_N_HEAP) != count + 1) {

-		fprintf(stderr, "N heap is wrong %lu %lu in index %s\n",
+		fprintf(stderr,
+		"InnoDB: N heap is wrong %lu %lu in index %s\n",
 			page_header_get_field(page, PAGE_N_HEAP), count + 1,
 				index->name);
+		goto func_exit;
 	}

 	ret = TRUE;	
@ -1532,6 +1743,15 @@ page_validate(
 func_exit:
 	mem_heap_free(heap);

+	if (ret == FALSE) {
+		fprintf(stderr,
+"InnoDB: Apparent corruption in page %lu in index %s in table %s\n",
+			buf_frame_get_page_no(page), index->name,
+			index->table_name);
+
+		buf_page_print(page);
+	}
+	
 	return(ret);			  
 }

--- a/innobase/pars/lexyy.c
+++ b/innobase/pars/lexyy.c
@ -4,8 +4,6 @@
 * $Header: /home/daffy/u0/vern/flex/RCS/flex.skl,v 2.91 96/09/10 16:58:48 vern Exp $
 */

-#include "univ.i"
-
 #define FLEX_SCANNER
 #define YY_FLEX_MAJOR_VERSION 2
 #define YY_FLEX_MINOR_VERSION 5
@ -609,11 +607,8 @@ How to make the InnoDB parser and lexer C files:

 6. Remove the #include of unistd.h from about line 2500 of lexyy.c

-7. Move #include <math.h> in pars0grm.c after #include "univ.i" to remove
-   a large file compilation error on AIX.
-
-8. Move #include "univ.i" in lexyy.c to the file start to remove a large
-   file compilation error on AIX.
+7. Add '#include "univ.i"' before #include <stdio.h> in lexyy.c
+   (Needed for AIX)

 These instructions seem to work at least with bison-1.28 and flex-2.5.4 on
 Linux.
@ -621,6 +616,7 @@ Linux.
 #line 36 "pars0lex.l"
 #define YYSTYPE que_node_t*

+#include "univ.i"
 #include "pars0pars.h"
 #include "pars0grm.h"
 #include "pars0sym.h"
--- a/innobase/pars/pars0grm.c
+++ b/innobase/pars/pars0grm.c
@ -102,8 +102,6 @@ que_node_t */
 #include "que0que.h"
 #include "row0sel.h"

-#include <math.h>
-
 #define YYSTYPE que_node_t*

 /* #define __STDC__ */
--- a/innobase/que/que0que.c
+++ b/innobase/que/que0que.c
@ -1046,14 +1046,16 @@ que_thr_stop(
 }

 /**************************************************************************
-A patch for MySQL used to 'stop' a dummy query thread used in MySQL. */
+A patch for MySQL used to 'stop' a dummy query thread used in MySQL. The
+query thread is stopped and made inactive, except in the case where
+it was put to the lock wait state in lock0lock.c, but the lock has already
+been granted or the transaction chosen as a victim in deadlock resolution. */

 void
 que_thr_stop_for_mysql(
 /*===================*/
 	que_thr_t*	thr)	/* in: query thread */
 {
-	ibool	stopped 	= FALSE;
 	trx_t*	trx;

 	trx = thr_get_trx(thr);
@ -1067,13 +1069,10 @@ que_thr_stop_for_mysql(

 			/* Error handling built for the MySQL interface */
 			thr->state = QUE_THR_COMPLETED;
-
-			stopped = TRUE;
-		}
-		
-		if (!stopped) {
-			/* It must have been a lock wait but the
-			lock was already released */
+		} else {
+			/* It must have been a lock wait but the lock was
+			already released, or this transaction was chosen
+			as a victim in selective deadlock resolution */

 			mutex_exit(&kernel_mutex);

@ -1081,6 +1080,10 @@ que_thr_stop_for_mysql(
 		}
 	}
 		
+	ut_ad(thr->is_active == TRUE);
+	ut_ad(trx->n_active_thrs == 1);
+	ut_ad(thr->graph->n_active_thrs == 1);
+
 	thr->is_active = FALSE;
 	(thr->graph)->n_active_thrs--;

@ -1132,6 +1135,9 @@ que_thr_stop_for_mysql_no_error(
 	trx_t*		trx)	/* in: transaction */
 {
 	ut_ad(thr->state == QUE_THR_RUNNING);
+	ut_ad(thr->is_active == TRUE);
+	ut_ad(trx->n_active_thrs == 1);
+	ut_ad(thr->graph->n_active_thrs == 1);
 		
 	if (thr->magic_n != QUE_THR_MAGIC_N) {
 		fprintf(stderr,
--- a/innobase/read/read0read.c
+++ b/innobase/read/read0read.c
@ -200,6 +200,28 @@ read_view_close(
 	UT_LIST_REMOVE(view_list, trx_sys->view_list, view);
 } 

+/*************************************************************************
+Closes a consistent read view for MySQL. This function is called at an SQL
+statement end if the trx isolation level is <= TRX_ISO_READ_COMMITTED. */
+
+void
+read_view_close_for_mysql(
+/*======================*/
+	trx_t*	trx)	/* in: trx which has a read view */
+{
+	ut_a(trx->read_view);
+
+	mutex_enter(&kernel_mutex);
+
+	read_view_close(trx->read_view);
+
+	mem_heap_empty(trx->read_view_heap);
+
+	trx->read_view = NULL;
+
+	mutex_exit(&kernel_mutex);
+}
+	
 /*************************************************************************
 Prints a read view to stderr. */

--- a/innobase/rem/rem0cmp.c
+++ b/innobase/rem/rem0cmp.c
@ -353,7 +353,7 @@ cmp_data_data_slow(
 		data2++;
 	}

-	return(0);
+	return(0);		/* Not reached */
 }

 /*****************************************************************
--- a/innobase/row/row0ins.c
+++ b/innobase/row/row0ins.c
@ -321,59 +321,6 @@ row_ins_clust_index_entry_by_modify(
 	return(err);
 }

-/*******************************************************************
-Checks if a unique key violation to rec would occur at the index entry
-insert. */
-static
-ibool
-row_ins_dupl_error_with_rec(
-/*========================*/
-				/* out: TRUE if error */
-	rec_t*		rec,	/* in: user record; NOTE that we assume
-				that the caller already has a record lock on
-				the record! */
-	dtuple_t*	entry,	/* in: entry to insert */
-	dict_index_t*	index)	/* in: index */
-{
-	ulint	matched_fields;
-	ulint	matched_bytes;
-	ulint	n_unique;
-	ulint   i;
-	
-	n_unique = dict_index_get_n_unique(index);
-
-	matched_fields = 0;
-	matched_bytes = 0;
-
-	cmp_dtuple_rec_with_match(entry, rec, &matched_fields, &matched_bytes);
-
-	if (matched_fields < n_unique) {
-
-	        return(FALSE);
-	}
-
-	/* In a unique secondary index we allow equal key values if they
-	contain SQL NULLs */
-
-	if (!(index->type & DICT_CLUSTERED)) {
-
-	        for (i = 0; i < n_unique; i++) {
-	                if (UNIV_SQL_NULL == dfield_get_len(
-                                         dtuple_get_nth_field(entry, i))) {
-
-	                        return(FALSE);
-	                }
-	        }
-	}
-
-	if (!rec_get_deleted_flag(rec)) {
-
-	        return(TRUE);
-	}
-
-	return(FALSE);
-}	
-
 /*************************************************************************
 Either deletes or sets the referencing columns SQL NULL in a child row.
 Used in ON DELETE ... clause for foreign keys when a parent row is
@ -533,8 +480,12 @@ row_ins_foreign_delete_or_set_null(
 	err = lock_table(0, table, LOCK_IX, thr);

 	if (err == DB_SUCCESS) {
+		/* Here it suffices to use a LOCK_REC_NOT_GAP type lock;
+		we already have a normal shared lock on the appropriate
+		gap if the search criterion was not unique */
+		
 		err = lock_clust_rec_read_check_and_lock(0, clust_rec,
-						clust_index, LOCK_X, thr);
+				clust_index, LOCK_X, LOCK_REC_NOT_GAP, thr);
 	}
 	
 	if (err != DB_SUCCESS) {
@ -630,12 +581,14 @@ nonstandard_exit_func:

 /*************************************************************************
 Sets a shared lock on a record. Used in locking possible duplicate key
-records. */
+records and also in checking foreign key constraints. */
 static
 ulint
 row_ins_set_shared_rec_lock(
 /*========================*/
 				/* out: DB_SUCCESS or error code */
+	ulint		type, 	/* in: LOCK_ORDINARY, LOCK_GAP, or
+				LOCK_REC_NOT_GAP type lock */
 	rec_t*		rec,	/* in: record */
 	dict_index_t*	index,	/* in: index */
 	que_thr_t*	thr)	/* in: query thread */	
@ -644,10 +597,10 @@ row_ins_set_shared_rec_lock(

 	if (index->type & DICT_CLUSTERED) {
 		err = lock_clust_rec_read_check_and_lock(0, rec, index, LOCK_S,
-									thr);
+								type, thr);
 	} else {
 		err = lock_sec_rec_read_check_and_lock(0, rec, index, LOCK_S,
-									thr);
+								type, thr);
 	}

 	return(err);
@ -656,7 +609,7 @@ row_ins_set_shared_rec_lock(
 /*******************************************************************
 Checks if foreign key constraint fails for an index entry. Sets shared locks
 which lock either the success or the failure of the constraint. NOTE that
-the caller must have a shared latch on dict_foreign_key_check_lock. */
+the caller must have a shared latch on dict_operation_lock. */

 ulint
 row_ins_check_foreign_constraint(
@ -679,7 +632,7 @@ row_ins_check_foreign_constraint(
 	dict_table_t*	check_table;
 	dict_index_t*	check_index;
 	ulint		n_fields_cmp;
-	ibool           timeout_expired;
+	ibool		unique_search;
 	rec_t*		rec;
 	btr_pcur_t	pcur;
 	ibool		moved;
@ -689,7 +642,9 @@ row_ins_check_foreign_constraint(
 	mtr_t		mtr;

 run_again:
-	ut_ad(rw_lock_own(&dict_foreign_key_check_lock, RW_LOCK_SHARED));
+	ut_ad(rw_lock_own(&dict_operation_lock, RW_LOCK_SHARED));
+	
+	err = DB_SUCCESS;

 	if (thr_get_trx(thr)->check_foreigns == FALSE) {
 		/* The user has suppressed foreign key checks currently for
@ -748,6 +703,14 @@ run_again:

 	dtuple_set_n_fields_cmp(entry, foreign->n_fields);

+	if (dict_index_get_n_unique(check_index) <= foreign->n_fields) {
+		/* We can just set a LOCK_REC_NOT_GAP type lock */
+	
+		unique_search = TRUE;
+	} else {
+		unique_search = FALSE;
+	}
+
 	btr_pcur_open(check_index, entry, PAGE_CUR_GE,
 					BTR_SEARCH_LEAF, &pcur, &mtr);

@ -761,26 +724,46 @@ run_again:
 			goto next_rec;
 		}
 		
-		/* Try to place a lock on the index record */
-
-		err = row_ins_set_shared_rec_lock(rec, check_index, thr);
+		if (rec == page_get_supremum_rec(buf_frame_align(rec))) {
 		
+			err = row_ins_set_shared_rec_lock(LOCK_ORDINARY, rec,
+							check_index, thr);
 			if (err != DB_SUCCESS) {

 				break;
 			}

-		if (rec == page_get_supremum_rec(buf_frame_align(rec))) {
-		
 			goto next_rec;
 		}

 		cmp = cmp_dtuple_rec(entry, rec);

 		if (cmp == 0) {
-			if (!rec_get_deleted_flag(rec)) {
+			if (rec_get_deleted_flag(rec)) {
+				err = row_ins_set_shared_rec_lock(LOCK_ORDINARY,
+							rec, check_index, thr);
+				if (err != DB_SUCCESS) {
+
+					break;
+				}
+			} else {
 				/* Found a matching record */
 				
+				if (unique_search) {
+					err = row_ins_set_shared_rec_lock(
+							LOCK_REC_NOT_GAP,
+							rec, check_index, thr);
+				} else {
+					err = row_ins_set_shared_rec_lock(
+							LOCK_ORDINARY,
+							rec, check_index, thr);
+				}
+				
+				if (err != DB_SUCCESS) {
+
+					break;
+				}
+
 /*				printf(
 "FOREIGN: Found matching record from %s %s\n",
 		check_index->table_name, check_index->name);
@ -807,6 +790,13 @@ run_again:
 		}

 		if (cmp < 0) {
+			err = row_ins_set_shared_rec_lock(LOCK_GAP,
+						rec, check_index, thr);
+			if (err != DB_SUCCESS) {
+
+				break;
+			}
+
 			if (check_ref) {			
 				err = DB_NO_REFERENCED_ROW;
 			} else {
@ -844,14 +834,14 @@ do_possible_lock_wait:

 		que_thr_stop_for_mysql(thr);

-		timeout_expired = srv_suspend_mysql_thread(thr);
+		srv_suspend_mysql_thread(thr);
 	
-		if (!timeout_expired) {
+		if (thr_get_trx(thr)->error_state == DB_SUCCESS) {

 		        goto run_again;
 		}

-		err = DB_LOCK_WAIT_TIMEOUT;
+		err = thr_get_trx(thr)->error_state;
 	}

 	return(err);
@ -890,21 +880,16 @@ row_ins_check_foreign_constraints(
 									trx);
 			}

-			if (!trx->has_dict_foreign_key_check_lock) {
+			if (0 == trx->dict_operation_lock_mode) {
 				got_s_lock = TRUE;

-				rw_lock_s_lock(&dict_foreign_key_check_lock);
-
-				trx->has_dict_foreign_key_check_lock = TRUE;
+				row_mysql_freeze_data_dictionary(trx);
 			}

 			err = row_ins_check_foreign_constraint(TRUE, foreign,
 						table, index, entry, thr);
 			if (got_s_lock) {
-
-				rw_lock_s_unlock(&dict_foreign_key_check_lock);	
-
-				trx->has_dict_foreign_key_check_lock = FALSE;
+				row_mysql_unfreeze_data_dictionary(trx);
 			}
 				
 			if (err != DB_SUCCESS) {
@ -918,6 +903,59 @@ row_ins_check_foreign_constraints(
 	return(DB_SUCCESS);
 }

+/*******************************************************************
+Checks if a unique key violation to rec would occur at the index entry
+insert. */
+static
+ibool
+row_ins_dupl_error_with_rec(
+/*========================*/
+				/* out: TRUE if error */
+	rec_t*		rec,	/* in: user record; NOTE that we assume
+				that the caller already has a record lock on
+				the record! */
+	dtuple_t*	entry,	/* in: entry to insert */
+	dict_index_t*	index)	/* in: index */
+{
+	ulint	matched_fields;
+	ulint	matched_bytes;
+	ulint	n_unique;
+	ulint   i;
+	
+	n_unique = dict_index_get_n_unique(index);
+
+	matched_fields = 0;
+	matched_bytes = 0;
+
+	cmp_dtuple_rec_with_match(entry, rec, &matched_fields, &matched_bytes);
+
+	if (matched_fields < n_unique) {
+
+	        return(FALSE);
+	}
+
+	/* In a unique secondary index we allow equal key values if they
+	contain SQL NULLs */
+
+	if (!(index->type & DICT_CLUSTERED)) {
+
+	        for (i = 0; i < n_unique; i++) {
+	                if (UNIV_SQL_NULL == dfield_get_len(
+                                         dtuple_get_nth_field(entry, i))) {
+
+	                        return(FALSE);
+	                }
+	        }
+	}
+
+	if (!rec_get_deleted_flag(rec)) {
+
+	        return(TRUE);
+	}
+
+	return(FALSE);
+}	
+
 /*******************************************************************
 Scans a unique non-clustered index at a given index entry to determine
 whether a uniqueness violation has occurred for the key value of the entry.
@ -978,7 +1016,8 @@ row_ins_scan_sec_index_for_duplicate(
 				
 		/* Try to place a lock on the index record */

-		err = row_ins_set_shared_rec_lock(rec, index, thr);
+		err = row_ins_set_shared_rec_lock(LOCK_ORDINARY, rec, index,
+									thr);

 		if (err != DB_SUCCESS) {

@ -1082,8 +1121,8 @@ row_ins_duplicate_error_in_clust(
 			sure that in roll-forward we get the same duplicate
 			errors as in original execution */
 		
-			err = row_ins_set_shared_rec_lock(rec, cursor->index,
-									thr);
+			err = row_ins_set_shared_rec_lock(LOCK_REC_NOT_GAP,
+						rec, cursor->index, thr);
 			if (err != DB_SUCCESS) {
 					
 				return(err);
@ -1105,8 +1144,8 @@ row_ins_duplicate_error_in_clust(

 		if (rec != page_get_supremum_rec(page)) {

-			err = row_ins_set_shared_rec_lock(rec, cursor->index,
-									thr);
+			err = row_ins_set_shared_rec_lock(LOCK_REC_NOT_GAP,
+						rec, cursor->index, thr);
 			if (err != DB_SUCCESS) {
 					
 				return(err);
--- a/innobase/row/row0mysql.c
+++ b/innobase/row/row0mysql.c
@ -27,6 +27,7 @@ Created 9/17/2000 Heikki Tuuri
 #include "lock0lock.h"
 #include "rem0cmp.h"
 #include "log0log.h"
+#include "btr0sea.h"

 /* A dummy variable used to fool the compiler */
 ibool	row_mysql_identically_false	= FALSE;
@ -197,13 +198,13 @@ row_mysql_handle_errors(
 				/* out: TRUE if it was a lock wait and
 				we should continue running the query thread */
 	ulint*		new_err,/* out: possible new error encountered in
-				rollback, or the old error which was
-				during the function entry */
+				lock wait, or if no new error, the value
+				of trx->error_state at the entry of this
+				function */
 	trx_t*		trx,	/* in: transaction */
 	que_thr_t*	thr,	/* in: query thread */
 	trx_savept_t*	savept)	/* in: savepoint or NULL */
 {
-	ibool	timeout_expired;
 	ulint	err;

 handle_new_error:
@ -240,11 +241,9 @@ handle_new_error:
 		/* MySQL will roll back the latest SQL statement */
 	} else if (err == DB_LOCK_WAIT) {

-		timeout_expired = srv_suspend_mysql_thread(thr);
-
-		if (timeout_expired) {
-			trx->error_state = DB_LOCK_WAIT_TIMEOUT;
+		srv_suspend_mysql_thread(thr);

+		if (trx->error_state != DB_SUCCESS) {
 			que_thr_stop_for_mysql(thr);

 			goto handle_new_error;
@ -321,6 +320,8 @@ row_create_prebuilt(

 	prebuilt->sql_stat_start = TRUE;

+	prebuilt->mysql_has_locked = FALSE;
+
 	prebuilt->index = NULL;
 	prebuilt->n_template = 0;
 	prebuilt->mysql_template = NULL;
@ -1012,10 +1013,27 @@ run_again:

 	err = trx->error_state;

+	/* Note that the cascade node is a subnode of another InnoDB
+	query graph node. We do a normal lock wait in this node, but
+	all errors are handled by the parent node. */
+
 	if (err == DB_LOCK_WAIT) {
+		/* Handle lock wait here */
+	
 		que_thr_stop_for_mysql(thr);

-		row_mysql_handle_errors(&err, trx, thr, NULL);
+		srv_suspend_mysql_thread(thr);
+
+		/* Note that a lock wait may also end in a lock wait timeout,
+		or this transaction is picked as a victim in selective
+		deadlock resolution */
+
+		if (trx->error_state != DB_SUCCESS) {
+
+			return(trx->error_state);
+		}
+
+		/* Retry operation after a normal lock wait */
 		
 		goto run_again;
 	}
@ -1136,32 +1154,73 @@ row_mysql_recover_tmp_table(
 }

 /*************************************************************************
-Locks the data dictionary exclusively for performing a table create
-operation. */
+Locks the data dictionary in shared mode from modifications, for performing
+foreign key check, rollback, or other operation invisible to MySQL. */

 void
-row_mysql_lock_data_dictionary(void)
-/*================================*/
+row_mysql_freeze_data_dictionary(
+/*=============================*/
+	trx_t*	trx)	/* in: transaction */
 {
+	ut_a(trx->dict_operation_lock_mode == 0);
+	
+	rw_lock_s_lock(&dict_operation_lock);
+
+	trx->dict_operation_lock_mode = RW_S_LATCH;
+}
+
+/*************************************************************************
+Unlocks the data dictionary shared lock. */
+
+void
+row_mysql_unfreeze_data_dictionary(
+/*===============================*/
+	trx_t*	trx)	/* in: transaction */
+{
+	ut_a(trx->dict_operation_lock_mode == RW_S_LATCH);
+
+	rw_lock_s_unlock(&dict_operation_lock);
+
+	trx->dict_operation_lock_mode = 0;
+}
+
+/*************************************************************************
+Locks the data dictionary exclusively for performing a table create or other
+data dictionary modification operation. */
+
+void
+row_mysql_lock_data_dictionary(
+/*===========================*/
+	trx_t*	trx)	/* in: transaction */
+{
+	ut_a(trx->dict_operation_lock_mode == 0);
+	
 	/* Serialize data dictionary operations with dictionary mutex:
 	no deadlocks or lock waits can occur then in these operations */

-	rw_lock_x_lock(&(dict_foreign_key_check_lock));
+	rw_lock_x_lock(&dict_operation_lock);
+	trx->dict_operation_lock_mode = RW_X_LATCH;
+
 	mutex_enter(&(dict_sys->mutex));
 }

 /*************************************************************************
-Unlocks the data dictionary exclusively lock. */
+Unlocks the data dictionary exclusive lock. */

 void
-row_mysql_unlock_data_dictionary(void)
-/*==================================*/
+row_mysql_unlock_data_dictionary(
+/*=============================*/
+	trx_t*	trx)	/* in: transaction */
 {
+	ut_a(trx->dict_operation_lock_mode == RW_X_LATCH);
+
 	/* Serialize data dictionary operations with dictionary mutex:
 	no deadlocks can occur then in these operations */

 	mutex_exit(&(dict_sys->mutex));
-	rw_lock_x_unlock(&(dict_foreign_key_check_lock));
+	rw_lock_x_unlock(&dict_operation_lock);
+
+	trx->dict_operation_lock_mode = 0;
 }

 /*************************************************************************
@ -1184,6 +1243,8 @@ row_create_table_for_mysql(
 	ulint		err;

 	ut_ad(trx->mysql_thread_id == os_thread_get_curr_id());
+	ut_ad(rw_lock_own(&dict_operation_lock, RW_LOCK_EX));
+	ut_ad(trx->dict_operation_lock_mode == RW_X_LATCH);
 	ut_ad(mutex_own(&(dict_sys->mutex)));
 	
 	if (srv_created_new_raw) {
@ -1332,7 +1393,7 @@ row_create_table_for_mysql(
 			fprintf(stderr, 
     "InnoDB: Warning: cannot create table %s because tablespace full\n",
 				 table->name);
-		     	row_drop_table_for_mysql(table->name, trx, TRUE);
+		     	row_drop_table_for_mysql(table->name, trx);
 		} else {
 		       	ut_a(err == DB_DUPLICATE_KEY);

@ -1384,6 +1445,7 @@ row_create_index_for_mysql(
 	ulint		keywordlen;
 	ulint		err;

+	ut_ad(rw_lock_own(&dict_operation_lock, RW_LOCK_EX));
 	ut_ad(mutex_own(&(dict_sys->mutex)));
 	ut_ad(trx->mysql_thread_id == os_thread_get_curr_id());
 	
@ -1425,7 +1487,7 @@ row_create_index_for_mysql(

 		trx_general_rollback_for_mysql(trx, FALSE, NULL);

-		row_drop_table_for_mysql(index->table_name, trx, TRUE);
+		row_drop_table_for_mysql(index->table_name, trx);

 		trx->error_state = DB_SUCCESS;
 	}
@ -1464,6 +1526,7 @@ row_table_add_foreign_constraints(
 	ulint	err;

 	ut_ad(mutex_own(&(dict_sys->mutex)));
+	ut_ad(rw_lock_own(&dict_operation_lock, RW_LOCK_EX));
 	ut_a(sql_string);
 	
 	trx->op_info = (char *) "adding foreign keys";
@ -1498,7 +1561,7 @@ row_table_add_foreign_constraints(

 		trx_general_rollback_for_mysql(trx, FALSE, NULL);

-		row_drop_table_for_mysql(name, trx, TRUE);
+		row_drop_table_for_mysql(name, trx);

 		trx->error_state = DB_SUCCESS;
 	}
@ -1529,7 +1592,7 @@ row_drop_table_for_mysql_in_background(
 							name); */
  	/* Drop the table in InnoDB */

-  	error = row_drop_table_for_mysql(name, trx, FALSE);
+  	error = row_drop_table_for_mysql(name, trx);

 	if (error != DB_SUCCESS) {
 		fprintf(stderr,
@ -1688,9 +1751,7 @@ row_drop_table_for_mysql(
 /*=====================*/
 				/* out: error code or DB_SUCCESS */
 	char*	name,		/* in: table name */
-	trx_t*	trx,		/* in: transaction handle */
-	ibool	has_dict_mutex)	/* in: TRUE if the caller already owns the
-				dictionary system mutex */
+	trx_t*	trx)		/* in: transaction handle */
 {
 	dict_table_t*	table;
 	que_thr_t*	thr;
@ -1702,6 +1763,7 @@ row_drop_table_for_mysql(
 	ulint		namelen;
 	ulint		keywordlen;
 	ulint		rounds	= 0;
+	ibool		locked_dictionary	= FALSE;
 	char		buf[10000];

 	ut_ad(trx->mysql_thread_id == os_thread_get_curr_id());
@ -1845,13 +1907,18 @@ row_drop_table_for_mysql(
 	/* Serialize data dictionary operations with dictionary mutex:
 	no deadlocks can occur then in these operations */

-	if (!has_dict_mutex) {
-		/* Prevent foreign key checks while we are dropping the table */
-		rw_lock_x_lock(&(dict_foreign_key_check_lock));
+	if (trx->dict_operation_lock_mode != RW_X_LATCH) {
+		/* Prevent foreign key checks etc. while we are dropping the
+		table */

-		mutex_enter(&(dict_sys->mutex));
+		row_mysql_lock_data_dictionary(trx);
+
+		locked_dictionary = TRUE;
 	}

+	ut_ad(mutex_own(&(dict_sys->mutex)));
+	ut_ad(rw_lock_own(&dict_operation_lock, RW_LOCK_EX));
+	
 	graph = pars_sql(buf);

 	ut_a(graph);
@ -1861,9 +1928,6 @@ row_drop_table_for_mysql(

 	graph->fork_type = QUE_FORK_MYSQL_INTERFACE;

-	/* Prevent purge from running while we are dropping the table */
-	rw_lock_s_lock(&(purge_sys->purge_is_running));
-
 	table = dict_table_get_low(name);

 	if (!table) {
@ -1945,11 +2009,9 @@ row_drop_table_for_mysql(
 		}
 	}
 funct_exit:
-	rw_lock_s_unlock(&(purge_sys->purge_is_running));

-	if (!has_dict_mutex) {
-		mutex_exit(&(dict_sys->mutex));
-		rw_lock_x_unlock(&(dict_foreign_key_check_lock));
+	if (locked_dictionary) {
+		row_mysql_unlock_data_dictionary(trx);	
 	}

 	que_graph_free(graph);
@ -1985,8 +2047,7 @@ row_drop_database_for_mysql(
 	
 	trx_start_if_not_started(trx);
 loop:
-	rw_lock_x_lock(&(dict_foreign_key_check_lock));
-	mutex_enter(&(dict_sys->mutex));
+	row_mysql_lock_data_dictionary(trx);

 	while ((table_name = dict_get_first_table_name_in_db(name))) {
 		ut_a(memcmp(table_name, name, strlen(name)) == 0);
@ -1999,8 +2060,7 @@ loop:
 		the table */

 		if (table->n_mysql_handles_opened > 0) {
-		        mutex_exit(&(dict_sys->mutex));
-			rw_lock_x_unlock(&(dict_foreign_key_check_lock));
+			row_mysql_unlock_data_dictionary(trx);

 			ut_print_timestamp(stderr);
 			fprintf(stderr,
@ -2015,7 +2075,7 @@ loop:
 		        goto loop;
 		}

-		err = row_drop_table_for_mysql(table_name, trx, TRUE);
+		err = row_drop_table_for_mysql(table_name, trx);

 		mem_free(table_name);

@ -2027,8 +2087,7 @@ loop:
 		}
 	}

-	mutex_exit(&(dict_sys->mutex));
-	rw_lock_x_unlock(&(dict_foreign_key_check_lock));
+	row_mysql_unlock_data_dictionary(trx);
 	
 	trx_commit_for_mysql(trx);

@ -2165,8 +2224,7 @@ row_rename_table_for_mysql(
 	/* Serialize data dictionary operations with dictionary mutex:
 	no deadlocks can occur then in these operations */

-	rw_lock_x_lock(&(dict_foreign_key_check_lock));
-	mutex_enter(&(dict_sys->mutex));
+	row_mysql_lock_data_dictionary(trx);

 	table = dict_table_get_low(old_name);

@ -2248,8 +2306,7 @@ row_rename_table_for_mysql(
 		}
 	}
 funct_exit:	
-	mutex_exit(&(dict_sys->mutex));
-	rw_lock_x_unlock(&(dict_foreign_key_check_lock));
+	row_mysql_unlock_data_dictionary(trx);

 	que_graph_free(graph);
 	
@ -2399,9 +2456,19 @@ row_check_table_for_mysql(
 	ulint		n_rows;
 	ulint		n_rows_in_table	= ULINT_UNDEFINED;
 	ulint		ret 		= DB_SUCCESS;
+	ulint		old_isolation_level;
 	
 	prebuilt->trx->op_info = (char *) "checking table";

+	old_isolation_level = prebuilt->trx->isolation_level;
+
+	/* We must run the index record counts at an isolation level
+	>= READ COMMITTED, because a dirty read can see a wrong number
+	of records in some index; to play safe, we use always
+	REPEATABLE READ here */
+
+	prebuilt->trx->isolation_level = TRX_ISO_REPEATABLE_READ;
+	
 	index = dict_table_get_first_index(table);

 	while (index != NULL) {
@ -2433,6 +2500,9 @@ row_check_table_for_mysql(
 		index = dict_table_get_next_index(index);
 	}

+	/* Restore the original isolation level */
+	prebuilt->trx->isolation_level = old_isolation_level;
+	
 	/* We validate also the whole adaptive hash index for all tables
 	at every CHECK TABLE */

--- a/innobase/row/row0purge.c
+++ b/innobase/row/row0purge.c
@ -24,6 +24,7 @@ Created 3/14/1997 Heikki Tuuri
 #include "row0row.h"
 #include "row0upd.h"
 #include "row0vers.h"
+#include "row0mysql.h"
 #include "log0log.h"

 /************************************************************************
@ -453,7 +454,9 @@ static
 ibool
 row_purge_parse_undo_rec(
 /*=====================*/
-				/* out: TRUE if purge operation required */
+				/* out: TRUE if purge operation required:
+				NOTE that then the CALLER must unfreeze
+				data dictionary! */
 	purge_node_t*	node,	/* in: row undo node */
 	ibool*		updated_extern,
 				/* out: TRUE if an externally stored field
@ -462,6 +465,7 @@ row_purge_parse_undo_rec(
 {
 	dict_index_t*	clust_index;
 	byte*		ptr;
+	trx_t*		trx;
 	dulint		undo_no;
 	dulint		table_id;
 	dulint		trx_id;
@ -472,6 +476,8 @@ row_purge_parse_undo_rec(
 	
 	ut_ad(node && thr);

+	trx = thr_get_trx(thr);
+	
 	ptr = trx_undo_rec_get_pars(node->undo_rec, &type, &cmpl_info,
 					updated_extern, &undo_no, &table_id);
 	node->rec_type = type;
@ -493,18 +499,21 @@ row_purge_parse_undo_rec(
 	    	return(FALSE);
 	}
 	
+	/* Prevent DROP TABLE etc. from running when we are doing the purge
+	for this row */
+
+	row_mysql_freeze_data_dictionary(trx);
+
 	mutex_enter(&(dict_sys->mutex));

 	node->table = dict_table_get_on_id_low(table_id, thr_get_trx(thr));
 	
-	rw_lock_x_lock(&(purge_sys->purge_is_running));
-
 	mutex_exit(&(dict_sys->mutex));

 	if (node->table == NULL) {
 		/* The table has been dropped: no need to do purge */

-		rw_lock_x_unlock(&(purge_sys->purge_is_running));
+		row_mysql_unfreeze_data_dictionary(trx);

 		return(FALSE);
 	}
@ -514,7 +523,7 @@ row_purge_parse_undo_rec(
 	if (clust_index == NULL) {
 		/* The table was corrupt in the data dictionary */

-		rw_lock_x_unlock(&(purge_sys->purge_is_running));
+		row_mysql_unfreeze_data_dictionary(trx);

 		return(FALSE);
 	}
@ -552,9 +561,12 @@ row_purge(
 	dulint	roll_ptr;
 	ibool	purge_needed;
 	ibool	updated_extern;
+	trx_t*	trx;
 	
 	ut_ad(node && thr);

+	trx = thr_get_trx(thr);
+	
 	node->undo_rec = trx_purge_fetch_next_rec(&roll_ptr,
 						&(node->reservation),
 						node->heap);
@ -573,6 +585,8 @@ row_purge(
 	} else {
 		purge_needed = row_purge_parse_undo_rec(node, &updated_extern,
 									thr);
+		/* If purge_needed == TRUE, we must also remember to unfreeze
+		data dictionary! */
 	}

 	if (purge_needed) {
@ -594,7 +608,7 @@ row_purge(
 			btr_pcur_close(&(node->pcur));
 		}

-		rw_lock_x_unlock(&(purge_sys->purge_is_running));		
+		row_mysql_unfreeze_data_dictionary(trx);
 	}

 	/* Do some cleanup */
--- a/innobase/row/row0sel.c
+++ b/innobase/row/row0sel.c
@ -606,7 +606,7 @@ row_sel_get_clust_rec(
 		/* Try to place a lock on the index record */
 		
 		err = lock_clust_rec_read_check_and_lock(0, clust_rec, index,
-						node->row_lock_mode, thr);
+					node->row_lock_mode, LOCK_ORDINARY, thr);
 		if (err != DB_SUCCESS) {

 			return(err);
@ -678,16 +678,17 @@ sel_set_rec_lock(
 	rec_t*		rec,	/* in: record */
 	dict_index_t*	index,	/* in: index */
 	ulint		mode,	/* in: lock mode */
+	ulint		type, 	/* in: LOCK_ORDINARY, LOCK_GAP, or LOC_REC_NOT_GAP */
 	que_thr_t*	thr)	/* in: query thread */	
 {
 	ulint	err;

 	if (index->type & DICT_CLUSTERED) {
 		err = lock_clust_rec_read_check_and_lock(0, rec, index, mode,
-									thr);
+							type, thr);
 	} else {
 		err = lock_sec_rec_read_check_and_lock(0, rec, index, mode,
-									thr);
+							type, thr);
 	}

 	return(err);
@ -1154,7 +1155,7 @@ rec_loop:
 		
 		if (!consistent_read) {
 			err = sel_set_rec_lock(page_rec_get_next(rec), index,
-						node->row_lock_mode, thr);
+				node->row_lock_mode, LOCK_ORDINARY, thr);
 			if (err != DB_SUCCESS) {
 				/* Note that in this case we will store in pcur
 				the PREDECESSOR of the record we are waiting
@ -1180,8 +1181,8 @@ rec_loop:
 	if (!consistent_read) {
 		/* Try to place a lock on the index record */	

-		err = sel_set_rec_lock(rec, index, node->row_lock_mode, thr);
-
+		err = sel_set_rec_lock(rec, index, node->row_lock_mode,
+						LOCK_ORDINARY, thr);
 		if (err != DB_SUCCESS) {

 			goto lock_wait_or_error;
@ -2200,6 +2201,7 @@ row_sel_get_clust_rec_for_mysql(
 	rec_t*		old_vers;
 	ulint		err;
 	trx_t*		trx;
+	char		err_buf[1000];

 	*out_rec = NULL;
 	
@ -2213,14 +2215,41 @@ row_sel_get_clust_rec_for_mysql(

 	clust_rec = btr_pcur_get_rec(prebuilt->clust_pcur);

-	ut_ad(page_rec_is_user_rec(clust_rec));
+	if (!page_rec_is_user_rec(clust_rec)) {
+		ut_print_timestamp(stderr);
+	  	fprintf(stderr,
+		"  InnoDB: error clustered record for sec rec not found\n"
+		"InnoDB: index %s table %s\n", sec_index->name,
+		  	sec_index->table->name);
+
+	  	rec_sprintf(err_buf, 900, rec);
+	  	fprintf(stderr, "InnoDB: sec index record %s\n", err_buf);
+
+	  	rec_sprintf(err_buf, 900, clust_rec);
+	  	fprintf(stderr, "InnoDB: clust index record %s\n", err_buf);
+
+		trx = thr_get_trx(thr);
+		trx_print(err_buf, trx);
+
+	  	fprintf(stderr,
+		"%s\nInnoDB: Make a detailed bug report and send it\n",
+							err_buf);
+	  	fprintf(stderr, "InnoDB: to mysql@lists.mysql.com\n");
+
+		clust_rec = NULL;
+
+		goto func_exit;
+	}

 	if (prebuilt->select_lock_type != LOCK_NONE) {
-		/* Try to place a lock on the index record */
+		/* Try to place a lock on the index record; we are searching
+		the clust rec with a unique condition, hence
+		we set a LOCK_REC_NOT_GAP type lock */
 		
 		err = lock_clust_rec_read_check_and_lock(0, clust_rec,
 					clust_index,
-					prebuilt->select_lock_type, thr);
+					prebuilt->select_lock_type,
+					LOCK_REC_NOT_GAP, thr);
 		if (err != DB_SUCCESS) {

 			return(err);
@ -2233,7 +2262,11 @@ row_sel_get_clust_rec_for_mysql(

 		old_vers = NULL;

-		if (!lock_clust_rec_cons_read_sees(clust_rec, clust_index,
+		/* If the isolation level allows reading of uncommitted data,
+		then we never look for an earlier version */
+
+		if (trx->isolation_level > TRX_ISO_READ_UNCOMMITTED
+		    && !lock_clust_rec_cons_read_sees(clust_rec, clust_index,
 							trx->read_view)) {

 			err = row_sel_build_prev_vers_for_mysql(
@ -2275,6 +2308,7 @@ row_sel_get_clust_rec_for_mysql(
 		}
 	}

+func_exit:
 	*out_rec = clust_rec;

 	if (prebuilt->select_lock_type == LOCK_X) {
@ -2407,7 +2441,7 @@ row_sel_push_cache_row_for_mysql(
 /*************************************************************************
 Tries to do a shortcut to fetch a clustered index record with a unique key,
 using the hash index if possible (not always). We assume that the search
-mode is PAGE_CUR_GE, it is a consistent read, trx has already a read view,
+mode is PAGE_CUR_GE, it is a consistent read, there is a read view in trx,
 btr search latch has been locked in S-mode. */
 static
 ulint
@ -2516,17 +2550,23 @@ row_search_for_mysql(
 	ibool		was_lock_wait;
 	ulint		ret;
 	ulint		shortcut;
+	ibool		unique_search			= FALSE;
 	ibool		unique_search_from_clust_index	= FALSE;
 	ibool		mtr_has_extra_clust_latch 	= FALSE;
 	ibool		moves_up 			= FALSE;
+	ibool		set_also_gap_locks		= TRUE;
+					/* if the query is a plain
+					locking SELECT, and the isolation
+					level is <= TRX_ISO_READ_COMMITTED,
+					then this is set to FALSE */
+	ibool		success;
 	ulint		cnt				= 0;
+	ulint		next_offs;
 	mtr_t		mtr;
 	
 	ut_ad(index && pcur && search_tuple);
 	ut_ad(trx->mysql_thread_id == os_thread_get_curr_id());
 		
-	ut_ad(sync_thread_levels_empty_gen(FALSE));
-	
 	if (prebuilt->magic_n != ROW_PREBUILT_ALLOCATED) {
 		fprintf(stderr,
 		"InnoDB: Error: trying to free a corrupt\n"
@ -2543,6 +2583,9 @@ row_search_for_mysql(
 	
 	printf("N tables locked %lu\n", trx->mysql_n_tables_locked);
 */
+	/*-------------------------------------------------------------*/
+	/* PHASE 1: Try to pop the row from the prefetch cache */
+
 	if (direction == 0) {
 		trx->op_info = (char *) "starting index read";
 	
@ -2608,18 +2651,35 @@ row_search_for_mysql(

 	mtr_start(&mtr);

-	/* Since we must release the search system latch when we retrieve an
-	externally stored field, we cannot use the adaptive hash index in a
-	search in the case the row may be long and there may be externally
-	stored fields */
+	/* In a search where at most one record in the index may match, we
+	can use a LOCK_REC_NOT_GAP type record lock when locking a non-delete
+	marked matching record.
+
+	Note that in a unique secondary index there may be different delete
+	marked versions of a record where only the primary key values differ:
+	thus in a secondary index we must use next-key locks when locking
+	delete marked records. */
 	
 	if (match_mode == ROW_SEL_EXACT
 	    && index->type & DICT_UNIQUE
-		&& index->type & DICT_CLUSTERED
-		&& !prebuilt->templ_contains_blob
-		&& (prebuilt->mysql_row_len < UNIV_PAGE_SIZE / 8)
 	    && dtuple_get_n_fields(search_tuple)
 				== dict_index_get_n_unique(index)) {
+		unique_search = TRUE;
+	}
+
+	/*-------------------------------------------------------------*/
+	/* PHASE 2: Try fast adaptive hash index search if possible */
+
+	/* Next test if this is the special case where we can use the fast
+	adaptive hash index to try the search. Since we must release the
+	search system latch when we retrieve an externally stored field, we
+	cannot use the adaptive hash index in a search in the case the row
+	may be long and there may be externally stored fields */
+
+	if (unique_search	
+	    && index->type & DICT_CLUSTERED
+	    && !prebuilt->templ_contains_blob
+	    && (prebuilt->mysql_row_len < UNIV_PAGE_SIZE / 8)) {

 		if (direction == ROW_SEL_NEXT) {
 			/* MySQL sometimes seems to do fetch next even
@ -2642,8 +2702,9 @@ row_search_for_mysql(

 		unique_search_from_clust_index = TRUE;

-		if (trx->mysql_n_tables_locked == 0
-					&& !prebuilt->sql_stat_start) {
+		if (prebuilt->select_lock_type == LOCK_NONE
+		    && trx->isolation_level > TRX_ISO_READ_UNCOMMITTED
+		    && trx->read_view) {

 			/* This is a SELECT query done as a consistent read,
 			and the read view has already been allocated:
@ -2722,7 +2783,11 @@ row_search_for_mysql(
 			mtr_start(&mtr);
 		}
 	}
+
 no_shortcut:
+	/*-------------------------------------------------------------*/
+	/* PHASE 3: Open or restore index cursor position */
+
 	if (trx->has_search_latch) {
 		rw_lock_s_unlock(&btr_search_latch);
 		trx->has_search_latch = FALSE;
@ -2730,6 +2795,23 @@ no_shortcut:

 	trx_start_if_not_started(trx);

+	if (trx->isolation_level <= TRX_ISO_READ_COMMITTED
+	    && prebuilt->select_lock_type != LOCK_NONE
+	    && trx->mysql_query_str) {
+
+		/* Scan the MySQL query string; check if SELECT is the first
+	        word there */
+
+		dict_accept(*trx->mysql_query_str, "SELECT", &success);
+
+		if (success) {
+			/* It is a plain locking SELECT and the isolation
+			level is low: do not lock gaps */
+
+			set_also_gap_locks = FALSE;
+		}
+	}
+	
 	/* Note that if the search mode was GE or G, then the cursor
 	naturally moves upward (in fetch next) in alphabetical order,
 	otherwise downward */
@ -2793,8 +2875,10 @@ no_shortcut:
 		prebuilt->sql_stat_start = FALSE;
 	}

-	/*-------------------------------------------------------------*/
 rec_loop:
+	/*-------------------------------------------------------------*/
+	/* PHASE 4: Look for matching records in a loop */
+	
 	cons_read_requires_clust_rec = FALSE;

 	rec = btr_pcur_get_rec(pcur);
@ -2813,36 +2897,87 @@ rec_loop:
 		goto next_rec;
 	}

-	if (prebuilt->select_lock_type != LOCK_NONE) {
+	if (rec == page_get_supremum_rec(buf_frame_align(rec))) {
+
+		if (prebuilt->select_lock_type != LOCK_NONE
+		    && set_also_gap_locks) {
+
 			/* Try to place a lock on the index record */	

-		err = sel_set_rec_lock(rec, index, prebuilt->select_lock_type,
-									thr);
+			err = sel_set_rec_lock(rec, index,
+						prebuilt->select_lock_type,
+						LOCK_ORDINARY, thr);
 			if (err != DB_SUCCESS) {

 				goto lock_wait_or_error;
 			}
 		}
-
-	if (rec == page_get_supremum_rec(buf_frame_align(rec))) {
-
 		/* A page supremum record cannot be in the result set: skip
-		it now when we have placed a possible lock on it */		
+		it now that we have placed a possible lock on it */
 		
 		goto next_rec;
 	}

-	ut_ad(page_rec_is_user_rec(rec));
+	/*-------------------------------------------------------------*/
+	/* Do sanity checks in case our cursor has bumped into page
+	corruption */
 	
-	if (unique_search_from_clust_index && btr_pcur_get_up_match(pcur)
-					== dtuple_get_n_fields(search_tuple)) {
-		/* The record matches enough */
+	next_offs = rec_get_next_offs(rec);

-		ut_ad(mode == PAGE_CUR_GE);
-#ifdef UNIV_SEARCH_DEBUG
-		ut_a(0 == cmp_dtuple_rec(search_tuple, rec));
-#endif	
-	} else if (match_mode == ROW_SEL_EXACT) {
+	if (next_offs >= UNIV_PAGE_SIZE || next_offs < PAGE_SUPREMUM) {
+
+		if (srv_force_recovery == 0 || moves_up == FALSE) {
+			ut_print_timestamp(stderr);
+			fprintf(stderr,
+"  InnoDB: Index corruption: rec offs %lu next offs %lu, page no %lu,\n"
+"InnoDB: index %s, table %s. Run CHECK TABLE to table. You may need to\n"
+"InnoDB: restore from a backup, or dump + drop + reimport the table.\n",
+			   (ulint)(rec - buf_frame_align(rec)), next_offs,
+			   buf_frame_get_page_no(rec), index->name,
+			   index->table_name);
+		
+			err = DB_CORRUPTION;
+
+			goto lock_wait_or_error;
+		} else {
+			/* The user may be dumping a corrupt table. Jump
+			over the corruption to recover as much as possible. */
+
+			fprintf(stderr,
+"InnoDB: Index corruption: rec offs %lu next offs %lu, page no %lu,\n"
+"InnoDB: index %s, table %s. We try to skip the rest of the page.\n",
+			   (ulint)(rec - buf_frame_align(rec)), next_offs,
+			   buf_frame_get_page_no(rec), index->name,
+			   index->table_name);
+
+			btr_pcur_move_to_last_on_page(pcur, &mtr);
+
+			goto next_rec;
+		}
+	}
+
+	if (srv_force_recovery > 0) {
+		if (!rec_validate(rec) || !btr_index_rec_validate(rec, index,
+								FALSE)) {
+			fprintf(stderr,
+"InnoDB: Index record corruption: rec offs %lu next offs %lu, page no %lu,\n"
+"InnoDB: index %s, table %s. We try to skip the record.\n",
+			   (ulint)(rec - buf_frame_align(rec)), next_offs,
+			   buf_frame_get_page_no(rec), index->name,
+			   index->table_name);
+
+			goto next_rec;
+		}
+	}
+
+	/*-------------------------------------------------------------*/
+
+	/* Note that we cannot trust the up_match value in the cursor at this
+	place because we can arrive here after moving the cursor! Thus
+	we have to recompare rec and search_tuple to determine if they
+	match enough. */
+
+	if (match_mode == ROW_SEL_EXACT) {
 		/* Test if the index record matches completely to search_tuple
 		in prebuilt: if not, then we return with DB_RECORD_NOT_FOUND */

@ -2850,6 +2985,19 @@ rec_loop:
 		
 		if (0 != cmp_dtuple_rec(search_tuple, rec)) {

+			if (prebuilt->select_lock_type != LOCK_NONE
+		    	    && set_also_gap_locks) {
+				/* Try to place a lock on the index record */	
+
+				err = sel_set_rec_lock(rec, index,
+						prebuilt->select_lock_type,
+						LOCK_GAP, thr);
+				if (err != DB_SUCCESS) {
+
+					goto lock_wait_or_error;
+				}
+			}
+
 			btr_pcur_store_position(pcur, &mtr);

 			ret = DB_RECORD_NOT_FOUND;
@ -2862,6 +3010,19 @@ rec_loop:

 		if (!cmp_dtuple_is_prefix_of_rec(search_tuple, rec)) {
 			
+			if (prebuilt->select_lock_type != LOCK_NONE
+			    && set_also_gap_locks) {
+				/* Try to place a lock on the index record */	
+
+				err = sel_set_rec_lock(rec, index,
+						prebuilt->select_lock_type,
+						LOCK_GAP, thr);
+				if (err != DB_SUCCESS) {
+
+					goto lock_wait_or_error;
+				}
+			}
+
 			btr_pcur_store_position(pcur, &mtr);

 			ret = DB_RECORD_NOT_FOUND;
@ -2874,16 +3035,39 @@ rec_loop:
 	/* We are ready to look at a possible new index entry in the result
 	set: the cursor is now placed on a user record */

-	/* Get the right version of the row in a consistent read */
+	if (prebuilt->select_lock_type != LOCK_NONE) {
+		/* Try to place a lock on the index record; note that delete
+		marked records are a special case in a unique search. If there
+		is a non-delete marked record, then it is enough to lock its
+		existence with LOCK_REC_NOT_GAP. */

-	if (prebuilt->select_lock_type == LOCK_NONE) {
+		if (!set_also_gap_locks
+		    || (unique_search && !rec_get_deleted_flag(rec))) {
+			err = sel_set_rec_lock(rec, index,
+						prebuilt->select_lock_type,
+						LOCK_REC_NOT_GAP, thr);
+		} else {
+			err = sel_set_rec_lock(rec, index,
+						prebuilt->select_lock_type,
+						LOCK_ORDINARY, thr);
+		}
 		
+		if (err != DB_SUCCESS) {
+
+			goto lock_wait_or_error;
+		}
+	} else {
 		/* This is a non-locking consistent read: if necessary, fetch
 		a previous version of the record */

 		cons_read_requires_clust_rec = FALSE;

-		if (index == clust_index) {
+		if (trx->isolation_level == TRX_ISO_READ_UNCOMMITTED) {
+
+			/* Do nothing: we let a non-locking SELECT read the
+			latest version of the record */
+		
+		} else if (index == clust_index) {
 			
 			if (!lock_clust_rec_cons_read_sees(rec, index,
 							trx->read_view)) {
@ -3020,8 +3204,11 @@ got_row:
 	ret = DB_SUCCESS;

 	goto normal_return;
-	/*-------------------------------------------------------------*/	
+
 next_rec:
+	/*-------------------------------------------------------------*/	
+	/* PHASE 5: Move the cursor to the next index record */
+	
 	if (mtr_has_extra_clust_latch) {
 		/* We must commit mtr if we are moving to the next
 		non-clustered index record, because we could break the
@ -3064,8 +3251,10 @@ next_rec:
 	cnt++;

 	goto rec_loop;
-	/*-------------------------------------------------------------*/
+
 lock_wait_or_error:
+	/*-------------------------------------------------------------*/
+
 	btr_pcur_store_position(pcur, &mtr);

 	mtr_commit(&mtr);
@ -3096,6 +3285,7 @@ lock_wait_or_error:
 	return(err);

 normal_return:
+	/*-------------------------------------------------------------*/
 	que_thr_stop_for_mysql_no_error(thr, trx);

 	mtr_commit(&mtr);
@ -3156,10 +3346,12 @@ row_search_check_if_query_cache_permitted(

 		ret = TRUE;
 		
-		/* Assign a read view for the transaction if it does not yet
-		have one */
+		/* If the isolation level is high, assign a read view for the
+		transaction if it does not yet have one */
+
+		if (trx->isolation_level >= TRX_ISO_REPEATABLE_READ
+		    && !trx->read_view) {

-		if (!trx->read_view) {
 			trx->read_view = read_view_open_now(trx,
 						trx->read_view_heap);
 		}
--- a/innobase/row/row0uins.c
+++ b/innobase/row/row0uins.c
@ -254,6 +254,7 @@ row_undo_ins_parse_undo_rec(
 	node->table = dict_table_get_on_id(table_id, node->trx);

 	if (node->table == NULL) {
+
 		return;
 	}

@ -292,6 +293,7 @@ row_undo_ins(

 	if (!found) {
 	        trx_undo_rec_release(node->trx, node->undo_no);
+
 		return(DB_SUCCESS);
 	}

--- a/innobase/row/row0undo.c
+++ b/innobase/row/row0undo.c
@ -24,6 +24,7 @@ Created 1/8/1997 Heikki Tuuri
 #include "row0row.h"
 #include "row0uins.h"
 #include "row0umod.h"
+#include "row0mysql.h"
 #include "srv0srv.h"

 /* How to undo row operations?
@ -204,6 +205,7 @@ row_undo(
 	ulint	err;
 	trx_t*	trx;
 	dulint	roll_ptr;
+	ibool	froze_data_dict	= FALSE;
 	
 	ut_ad(node && thr);
 	
@ -211,7 +213,6 @@ row_undo(

 	if (node->state == UNDO_NODE_FETCH_NEXT) {

-		/* The call below also starts &mtr */
 		node->undo_rec = trx_roll_pop_top_rec_of_trx(trx,
 							trx->roll_limit,
 							&roll_ptr,
@ -254,6 +255,18 @@ row_undo(
 		}
 	}

+	/* Prevent DROP TABLE etc. while we are rolling back this row.
+        If we are doing a TABLE CREATE or some other dictionary operation,
+        then we already have dict_operation_lock locked in x-mode. Do not
+        try to lock again in s-mode, because that would cause a hang. */
+
+	if (trx->dict_operation_lock_mode == 0) {
+        
+	        row_mysql_freeze_data_dictionary(trx);
+
+	        froze_data_dict = TRUE;
+	}
+
 	if (node->state == UNDO_NODE_INSERT) {

 		err = row_undo_ins(node, thr);
@ -264,6 +277,11 @@ row_undo(
 		err = row_undo_mod(node, thr);
 	}

+	if (froze_data_dict) {
+
+	        row_mysql_unfreeze_data_dictionary(trx);
+	}
+
 	/* Do some cleanup */
 	btr_pcur_close(&(node->pcur));

--- a/innobase/row/row0upd.c
+++ b/innobase/row/row0upd.c
@ -79,7 +79,7 @@ ibool
 row_upd_index_is_referenced(
 /*========================*/
 				/* out: TRUE if referenced; NOTE that since
-				we do not hold dict_foreign_key_check_lock
+				we do not hold dict_operation_lock
 				when leaving the function, it may be that
 				the referencing table has been dropped when
 				we leave this function: this function is only
@ -89,14 +89,16 @@ row_upd_index_is_referenced(
 {
 	dict_table_t*	table		= index->table;
 	dict_foreign_t*	foreign;
+	ibool		froze_data_dict	= FALSE;

 	if (!UT_LIST_GET_FIRST(table->referenced_list)) {

 		return(FALSE);
 	}

-	if (!trx->has_dict_foreign_key_check_lock) {
-		rw_lock_s_lock(&dict_foreign_key_check_lock);
+	if (trx->dict_operation_lock_mode == 0) {
+		row_mysql_freeze_data_dictionary(trx);
+		froze_data_dict = TRUE;
 	}

 	foreign = UT_LIST_GET_FIRST(table->referenced_list);
@ -104,8 +106,8 @@ row_upd_index_is_referenced(
 	while (foreign) {
 		if (foreign->referenced_index == index) {

-			if (!trx->has_dict_foreign_key_check_lock) {
-				rw_lock_s_unlock(&dict_foreign_key_check_lock);
+			if (froze_data_dict) {
+				row_mysql_unfreeze_data_dictionary(trx);
 			}

 			return(TRUE);
@ -114,8 +116,8 @@ row_upd_index_is_referenced(
 		foreign = UT_LIST_GET_NEXT(referenced_list, foreign);
 	}
 	
-	if (!trx->has_dict_foreign_key_check_lock) {
-		rw_lock_s_unlock(&dict_foreign_key_check_lock);
+	if (froze_data_dict) {
+		row_mysql_unfreeze_data_dictionary(trx);
 	}

 	return(FALSE);
@ -162,12 +164,10 @@ row_upd_check_references_constraints(

 	mtr_start(mtr);	
 	
-	if (!trx->has_dict_foreign_key_check_lock) {
+	if (trx->dict_operation_lock_mode == 0) {
 		got_s_lock = TRUE;

-		rw_lock_s_lock(&dict_foreign_key_check_lock);
-
-		trx->has_dict_foreign_key_check_lock = TRUE;
+		row_mysql_freeze_data_dictionary(trx);
 	}
 		
 	foreign = UT_LIST_GET_FIRST(table->referenced_list);
@ -189,7 +189,7 @@ row_upd_check_references_constraints(
 			}

 			/* NOTE that if the thread ends up waiting for a lock
-			we will release dict_foreign_key_check_lock
+			we will release dict_operation_lock
 			temporarily! But the counter on the table
 			protects 'foreign' from being dropped while the check
 			is running. */
@ -211,10 +211,7 @@ row_upd_check_references_constraints(

 			if (err != DB_SUCCESS) {
 				if (got_s_lock) {
-					rw_lock_s_unlock(
-						&dict_foreign_key_check_lock);	
-					trx->has_dict_foreign_key_check_lock
-								= FALSE;
+					row_mysql_unfreeze_data_dictionary(trx);
 				}

 				mem_heap_free(heap);
@ -227,8 +224,7 @@ row_upd_check_references_constraints(
 	}

 	if (got_s_lock) {
-		rw_lock_s_unlock(&dict_foreign_key_check_lock);	
-		trx->has_dict_foreign_key_check_lock = FALSE;
+		row_mysql_unfreeze_data_dictionary(trx);
 	}

 	mem_heap_free(heap);
--- a/innobase/srv/srv0srv.c
+++ b/innobase/srv/srv0srv.c
@ -51,6 +51,10 @@ Created 10/8/1995 Heikki Tuuri
 #include "srv0start.h"
 #include "row0mysql.h"

+/* This is set to TRUE if the MySQL user has set it in MySQL; currently
+affects only FOREIGN KEY definition parsing */
+ibool	srv_lower_case_table_names	= FALSE;
+
 /* Buffer which can be used in printing fatal error messages */
 char	srv_fatal_errbuf[5000];

@ -136,8 +140,6 @@ byte	srv_latin1_ordering[256]	/* The sort order table of the latin1
 , 0xD8, 0x55, 0x55, 0x55, 0x59, 0x59, 0xDE, 0xFF
 };
 		
-ibool	srv_use_native_aio	= FALSE;
-		
 ulint	srv_pool_size		= ULINT_MAX;	/* size in database pages;
 						MySQL originally sets this
 						value in megabytes */ 
@ -151,8 +153,9 @@ dulint	srv_archive_recovery_limit_lsn;

 ulint	srv_lock_wait_timeout	= 1024 * 1024 * 1024;

-char*   srv_unix_file_flush_method_str = NULL;
-ulint   srv_unix_file_flush_method = 0;
+char*   srv_file_flush_method_str = NULL;
+ulint   srv_unix_file_flush_method = SRV_UNIX_FDATASYNC;
+ulint   srv_win_file_flush_method = SRV_WIN_IO_UNBUFFERED;

 /* If the following is != 0 we do not allow inserts etc. This protects
 the user from forgetting the innodb_force_recovery keyword to my.cnf */
@ -281,6 +284,9 @@ time_t	srv_last_monitor_time;

 mutex_t srv_innodb_monitor_mutex;

+ulint	srv_main_thread_process_no	= 0;
+ulint	srv_main_thread_id		= 0;
+
 /*
 	IMPLEMENTATION OF THE SERVER MAIN PROGRAM
 	=========================================
@ -2046,13 +2052,15 @@ srv_table_reserve_slot_for_mysql(void)
 }

 /*******************************************************************
-Puts a MySQL OS thread to wait for a lock to be released. */
+Puts a MySQL OS thread to wait for a lock to be released. If an error
+occurs during the wait trx->error_state associated with thr is
+!= DB_SUCCESS when we return. DB_LOCK_WAIT_TIMEOUT and DB_DEADLOCK
+are possible errors. DB_DEADLOCK is returned if selective deadlock
+resolution chose this transaction as a victim. */

-ibool
+void
 srv_suspend_mysql_thread(
 /*=====================*/
-				/* out: TRUE if the lock wait timeout was
-				exceeded */
 	que_thr_t*	thr)	/* in: query thread associated with the MySQL
 				OS thread */
 {
@ -2060,6 +2068,7 @@ srv_suspend_mysql_thread(
 	os_event_t	event;
 	double		wait_time;
 	trx_t*		trx;
+	ibool		had_dict_lock	= FALSE;
 	
 	ut_ad(!mutex_own(&kernel_mutex));

@ -2069,15 +2078,28 @@ srv_suspend_mysql_thread(

 	mutex_enter(&kernel_mutex);

+	trx->error_state = DB_SUCCESS;
+
 	if (thr->state == QUE_THR_RUNNING) {

-		/* The lock has already been released: no need to suspend */
+		ut_ad(thr->is_active == TRUE);
+	
+		/* The lock has already been released or this transaction
+		was chosen as a deadlock victim: no need to suspend */
+
+		if (trx->was_chosen_as_deadlock_victim) {
+
+			trx->error_state = DB_DEADLOCK;
+			trx->was_chosen_as_deadlock_victim = FALSE;
+		}

 		mutex_exit(&kernel_mutex);

-		return(FALSE);
+		return;
 	}
 	
+	ut_ad(thr->is_active == FALSE);
+
 	slot = srv_table_reserve_slot_for_mysql();

 	event = slot->event;
@ -2101,18 +2123,22 @@ srv_suspend_mysql_thread(
 	srv_conc_force_exit_innodb(thr_get_trx(thr));

 	/* Release possible foreign key check latch */
-	if (trx->has_dict_foreign_key_check_lock) {
+	if (trx->dict_operation_lock_mode == RW_S_LATCH) {

-		rw_lock_s_unlock(&dict_foreign_key_check_lock);
+		had_dict_lock = TRUE;
+
+		row_mysql_unfreeze_data_dictionary(trx);
 	}

+	ut_a(trx->dict_operation_lock_mode == 0);
+
 	/* Wait for the release */
 	
 	os_event_wait(event);

-	if (trx->has_dict_foreign_key_check_lock) {
+	if (had_dict_lock) {

-		rw_lock_s_lock(&dict_foreign_key_check_lock);
+		row_mysql_freeze_data_dictionary(trx);
 	}

 	/* Return back inside InnoDB */
@ -2127,14 +2153,19 @@ srv_suspend_mysql_thread(

 	wait_time = ut_difftime(ut_time(), slot->suspend_time);
 	
+	if (trx->was_chosen_as_deadlock_victim) {
+
+		trx->error_state = DB_DEADLOCK;
+		trx->was_chosen_as_deadlock_victim = FALSE;
+	}
+
 	mutex_exit(&kernel_mutex);

 	if (srv_lock_wait_timeout < 100000000 && 
 	    			wait_time > (double)srv_lock_wait_timeout) {
-	   	return(TRUE);
-	}

-	return(FALSE);
+	    	trx->error_state = DB_LOCK_WAIT_TIMEOUT;
+	}
 }

 /************************************************************************
@ -2300,9 +2331,19 @@ srv_sprintf_innodb_monitor(
 		       "ROW OPERATIONS\n"
 		       "--------------\n");
 	buf += sprintf(buf,
-	"%ld queries inside InnoDB, %ld queries in queue; main thread: %s\n",
-			srv_conc_n_threads, srv_conc_n_waiting_threads,
+	"%ld queries inside InnoDB, %ld queries in queue\n",
+			srv_conc_n_threads, srv_conc_n_waiting_threads);
+#ifdef UNIV_LINUX
+	buf += sprintf(buf,
+	"Main thread process no %lu, state: %s\n",
+			srv_main_thread_process_no,
 			srv_main_thread_op_info);
+#else
+	buf += sprintf(buf,
+	"Main thread id %lu, state: %s\n",
+			srv_main_thread_id,
+			srv_main_thread_op_info);
+#endif
 	buf += sprintf(buf,
 	"Number of rows inserted %lu, updated %lu, deleted %lu, read %lu\n",
 			srv_n_rows_inserted, 
@ -2636,6 +2677,9 @@ srv_master_thread(
 	
 	UT_NOT_USED(arg);

+	srv_main_thread_process_no = os_proc_get_number();
+	srv_main_thread_id = os_thread_pf(os_thread_get_curr_id());
+	
 	srv_table_reserve_slot(SRV_MASTER);	

 	mutex_enter(&kernel_mutex);
--- a/innobase/srv/srv0start.c
+++ b/innobase/srv/srv0start.c
@ -56,6 +56,10 @@ Created 2/16/1996 Heikki Tuuri
 #include "srv0start.h"
 #include "que0que.h"

+ibool           srv_start_has_been_called  = FALSE;
+
+ulint           srv_sizeof_trx_t_in_ha_innodb_cc;
+
 ibool           srv_startup_is_before_trx_rollback_phase = FALSE;
 ibool           srv_is_being_started = FALSE;
 ibool           srv_was_started      = FALSE;
@ -515,7 +519,7 @@ srv_calc_high32(
 }

 /*************************************************************************
-Creates or opens the log files. */
+Creates or opens the log files and closes them. */
 static
 ulint
 open_or_create_log_file(
@ -640,7 +644,7 @@ open_or_create_log_file(
 }

 /*************************************************************************
-Creates or opens database data files. */
+Creates or opens database data files and closes them. */
 static
 ulint
 open_or_create_data_files(
@ -960,36 +964,91 @@ innobase_start_or_create_for_mysql(void)
 "InnoDB: !!!!!!!!!!!!!! UNIV_MEM_DEBUG switched on !!!!!!!!!!!!!!!\n"); 
 #endif

+        if (srv_sizeof_trx_t_in_ha_innodb_cc != (ulint)sizeof(trx_t)) {
+	        fprintf(stderr,
+  "InnoDB: Error: trx_t size is %lu in ha_innodb.cc but %lu in srv0start.c\n"
+  "InnoDB: Check that pthread_mutex_t is defined in the same way in these\n"
+  "InnoDB: compilation modules. Cannot continue.\n",
+		  srv_sizeof_trx_t_in_ha_innodb_cc, (ulint)sizeof(trx_t));
+		return(DB_ERROR);
+	}
+
+	/* Since InnoDB does not currently clean up all its internal data
+	   structures in MySQL Embedded Server Library server_end(), we
+	   print an error message if someone tries to start up InnoDB a
+	   second time during the process lifetime. */
+
+	if (srv_start_has_been_called) {
+	        fprintf(stderr,
+"InnoDB: Error:startup called second time during the process lifetime.\n"
+"InnoDB: In the MySQL Embedded Server Library you cannot call server_init()\n"
+"InnoDB: more than once during the process lifetime.\n");
+	}
+
+	srv_start_has_been_called = TRUE;
+
 	log_do_write = TRUE;
 /*	yydebug = TRUE; */

 	srv_is_being_started = TRUE;
        srv_startup_is_before_trx_rollback_phase = TRUE;
+	os_aio_use_native_aio = FALSE;
+
+#ifdef __WIN__
+	if (os_get_os_version() == OS_WIN95
+	    || os_get_os_version() == OS_WIN31
+	    || os_get_os_version() == OS_WINNT) {
+
+	  	/* On Win 95, 98, ME, Win32 subsystem for Windows 3.1,
+		and NT use simulated aio. In NT Windows provides async i/o,
+		but when run in conjunction with InnoDB Hot Backup, it seemed
+		to corrupt the data files. */
+
+	  	os_aio_use_native_aio = FALSE;
+	} else {
+	  	/* On Win 2000 and XP use async i/o */
+	  	os_aio_use_native_aio = TRUE;
+	}
+#endif	
+        if (srv_file_flush_method_str == NULL) {
+        	/* These are the default options */

-	if (0 == ut_strcmp(srv_unix_file_flush_method_str, "fdatasync")) {
 		srv_unix_file_flush_method = SRV_UNIX_FDATASYNC;

-	} else if (0 == ut_strcmp(srv_unix_file_flush_method_str, "O_DSYNC")) {
+		srv_win_file_flush_method = SRV_WIN_IO_UNBUFFERED;
+#ifndef __WIN__        
+	} else if (0 == ut_strcmp(srv_file_flush_method_str, "fdatasync")) {
+	  	srv_unix_file_flush_method = SRV_UNIX_FDATASYNC;
+
+	} else if (0 == ut_strcmp(srv_file_flush_method_str, "O_DSYNC")) {
 	  	srv_unix_file_flush_method = SRV_UNIX_O_DSYNC;

-	} else if (0 == ut_strcmp(srv_unix_file_flush_method_str,
+	} else if (0 == ut_strcmp(srv_file_flush_method_str,
 				  "littlesync")) {
 	  	srv_unix_file_flush_method = SRV_UNIX_LITTLESYNC;

-	} else if (0 == ut_strcmp(srv_unix_file_flush_method_str, "nosync")) {
+	} else if (0 == ut_strcmp(srv_file_flush_method_str, "nosync")) {
 	  	srv_unix_file_flush_method = SRV_UNIX_NOSYNC;
+#else
+	} else if (0 == ut_strcmp(srv_file_flush_method_str, "normal")) {
+	  	srv_win_file_flush_method = SRV_WIN_IO_NORMAL;
+	  	os_aio_use_native_aio = FALSE;
+
+	} else if (0 == ut_strcmp(srv_file_flush_method_str, "unbuffered")) {
+	  	srv_win_file_flush_method = SRV_WIN_IO_UNBUFFERED;
+	  	os_aio_use_native_aio = FALSE;
+
+	} else if (0 == ut_strcmp(srv_file_flush_method_str,
+							"async_unbuffered")) {
+	  	srv_win_file_flush_method = SRV_WIN_IO_UNBUFFERED;	
+#endif
 	} else {
 	  	fprintf(stderr, 
          	"InnoDB: Unrecognized value %s for innodb_flush_method\n",
-          				srv_unix_file_flush_method_str);
+          				srv_file_flush_method_str);
 	  	return(DB_ERROR);
 	}

-	/*
-	printf("srv_unix set to %lu\n", srv_unix_file_flush_method);
-	*/
-	os_aio_use_native_aio = srv_use_native_aio;
-
 	err = srv_boot();

 	if (err != DB_SUCCESS) {
@ -999,34 +1058,15 @@ innobase_start_or_create_for_mysql(void)

 	/* Restrict the maximum number of file i/o threads */
 	if (srv_n_file_io_threads > SRV_MAX_N_IO_THREADS) {
+
 		srv_n_file_io_threads = SRV_MAX_N_IO_THREADS;
 	}

-#if !(defined(WIN_ASYNC_IO) || defined(POSIX_ASYNC_IO))
+	if (!os_aio_use_native_aio) {
 		/* In simulated aio we currently have use only for 4 threads */

-	os_aio_use_native_aio = FALSE;
-
 		srv_n_file_io_threads = 4;
-#endif

-#ifdef __WIN__
-	if (os_get_os_version() == OS_WIN95
-	    || os_get_os_version() == OS_WIN31) {
-
-	  	/* On Win 95, 98, ME, and Win32 subsystem for Windows 3.1 use
-	     	simulated aio */
-
-	  	os_aio_use_native_aio = FALSE;
-	  	srv_n_file_io_threads = 4;
-	} else {
-	  	/* On NT and Win 2000 always use aio */
-	  	os_aio_use_native_aio = TRUE;
-	}
-#endif
-	os_aio_use_native_aio = FALSE;
-	
-	if (!os_aio_use_native_aio) {
 		os_aio_init(8 * SRV_N_PENDING_IOS_PER_THREAD
 						* srv_n_file_io_threads,
 					srv_n_file_io_threads,
@ -1047,15 +1087,6 @@ innobase_start_or_create_for_mysql(void)
 	
 	lock_sys_create(srv_lock_table_size);

-#ifdef POSIX_ASYNC_IO
-	if (os_aio_use_native_aio) {
-		/* There is only one thread per async io array:
-		one for ibuf i/o, one for log i/o, one for ordinary reads,
-		one for ordinary writes; we need only 4 i/o threads */
-
-		srv_n_file_io_threads = 4;
-	}
-#endif
 	/* Create i/o-handler threads: */

 	for (i = 0; i < srv_n_file_io_threads; i++) {
@ -1365,7 +1396,7 @@ innobase_start_or_create_for_mysql(void)
 	if (0 != os_fast_mutex_trylock(&srv_os_test_mutex)) {
 	        fprintf(stderr,
 "InnoDB: Error: pthread_mutex_trylock returns an unexpected value on\n"
-		  "InnoDB: success! Cannot continue.\n");
+"InnoDB: success! Cannot continue.\n");
 	        exit(1);
 	}

@ -1375,11 +1406,17 @@ innobase_start_or_create_for_mysql(void)

 	os_fast_mutex_unlock(&srv_os_test_mutex);

-	if (srv_print_verbose_log)
-	{
+	if (srv_print_verbose_log) {
 	  	ut_print_timestamp(stderr);
 	  	fprintf(stderr, "  InnoDB: Started\n");
 	}
+
+	if (srv_force_recovery > 0) {
+		fprintf(stderr,
+		"InnoDB: !!! innodb_force_recovery is set to %lu !!!\n",
+			srv_force_recovery);
+	}	
+
 	return((int) DB_SUCCESS);
 }

--- a/innobase/stamp-h.in
+++ b/innobase/stamp-h.in
@ -1 +0,0 @@
-timestamp
--- a/innobase/sync/sync0rw.c
+++ b/innobase/sync/sync0rw.c
@ -663,7 +663,8 @@ rw_lock_own(
 /*========*/
 					/* out: TRUE if locked */
 	rw_lock_t*	lock,		/* in: rw-lock */
-	ulint		lock_type)	/* in: lock type */
+	ulint		lock_type)	/* in: lock type: RW_LOCK_SHARED,
+					RW_LOCK_EX */
 {
 	rw_lock_debug_t*	info;

--- a/innobase/sync/sync0sync.c
+++ b/innobase/sync/sync0sync.c
@ -901,8 +901,7 @@ sync_thread_levels_empty_gen(

 		if (slot->latch != NULL && (!dict_mutex_allowed ||
 				(slot->level != SYNC_DICT
-				&& slot->level != SYNC_FOREIGN_KEY_CHECK
-				&& slot->level != SYNC_PURGE_IS_RUNNING))) {
+				&& slot->level != SYNC_DICT_OPERATION))) {

 			lock = slot->latch;
 			mutex = slot->latch;
@ -1087,12 +1086,10 @@ sync_thread_add_level(
 						SYNC_IBUF_PESS_INSERT_MUTEX));
 	} else if (level == SYNC_DICT_AUTOINC_MUTEX) {
 		ut_a(sync_thread_levels_g(array, SYNC_DICT_AUTOINC_MUTEX));
-	} else if (level == SYNC_FOREIGN_KEY_CHECK) {
-		ut_a(sync_thread_levels_g(array, SYNC_FOREIGN_KEY_CHECK));
+	} else if (level == SYNC_DICT_OPERATION) {
+		ut_a(sync_thread_levels_g(array, SYNC_DICT_OPERATION));
 	} else if (level == SYNC_DICT_HEADER) {
 		ut_a(sync_thread_levels_g(array, SYNC_DICT_HEADER));
-	} else if (level == SYNC_PURGE_IS_RUNNING) {
-		ut_a(sync_thread_levels_g(array, SYNC_PURGE_IS_RUNNING));
 	} else if (level == SYNC_DICT) {
 		ut_a(buf_debug_prints
 		     || sync_thread_levels_g(array, SYNC_DICT));
--- a/innobase/trx/trx0purge.c
+++ b/innobase/trx/trx0purge.c
@ -209,9 +209,6 @@ trx_purge_sys_create(void)
 	purge_sys->purge_undo_no = ut_dulint_zero;
 	purge_sys->next_stored = FALSE;
 	
-	rw_lock_create(&(purge_sys->purge_is_running));
-	rw_lock_set_level(&(purge_sys->purge_is_running),
-						SYNC_PURGE_IS_RUNNING);
 	rw_lock_create(&(purge_sys->latch));
 	rw_lock_set_level(&(purge_sys->latch), SYNC_PURGE_LATCH);

--- a/innobase/trx/trx0roll.c
+++ b/innobase/trx/trx0roll.c
@ -254,7 +254,7 @@ loop:
 	mutex_exit(&kernel_mutex);

 	if (trx->dict_operation) {
-		mutex_enter(&(dict_sys->mutex));
+		row_mysql_lock_data_dictionary(trx);
 	}

 	que_run_threads(thr);
@ -290,14 +290,14 @@ loop:
 			fprintf(stderr,
 "InnoDB: Table found: dropping table %s in recovery\n", table->name);

-			err = row_drop_table_for_mysql(table->name, trx,
-								TRUE);
+			err = row_drop_table_for_mysql(table->name, trx);
+
 			ut_a(err == (int) DB_SUCCESS);
 		}
 	}

 	if (trx->dict_operation) {
-		mutex_exit(&(dict_sys->mutex));
+		row_mysql_unlock_data_dictionary(trx);
 	}

 	fprintf(stderr, "InnoDB: Rolling back of trx id %lu %lu completed\n",
--- a/innobase/trx/trx0trx.c
+++ b/innobase/trx/trx0trx.c
@ -23,7 +23,7 @@ Created 3/26/1996 Heikki Tuuri
 #include "srv0srv.h"
 #include "thr0loc.h"
 #include "btr0sea.h"
-
+#include "os0proc.h"

 /* Copy of the prototype for innobase_mysql_print_thd: this
 copy MUST be equal to the one in mysql/sql/ha_innobase.cc ! */
@ -85,12 +85,14 @@ trx_create(
 	trx->conc_state = TRX_NOT_STARTED;
 	trx->start_time = time(NULL);

+	trx->isolation_level = TRX_ISO_REPEATABLE_READ;
 	trx->check_foreigns = TRUE;
 	trx->check_unique_secondary = TRUE;

 	trx->dict_operation = FALSE;

 	trx->mysql_thd = NULL;
+	trx->mysql_query_str = NULL;

 	trx->n_mysql_tables_in_use = 0;
 	trx->mysql_n_tables_locked = 0;
@ -127,12 +129,13 @@ trx_create(
 	trx->graph = NULL;

 	trx->wait_lock = NULL;
+	trx->was_chosen_as_deadlock_victim = FALSE;
 	UT_LIST_INIT(trx->wait_thrs);

 	trx->lock_heap = mem_heap_create_in_buffer(256);
 	UT_LIST_INIT(trx->trx_locks);

-	trx->has_dict_foreign_key_check_lock = FALSE;
+	trx->dict_operation_lock_mode = 0;
 	trx->has_search_latch = FALSE;
 	trx->search_latch_timeout = BTR_SEA_TIMEOUT;

@ -176,6 +179,8 @@ trx_allocate_for_mysql(void)

 	trx->mysql_thread_id = os_thread_get_curr_id();

+	trx->mysql_process_no = os_proc_get_number();
+	
 	return(trx);
 }

@ -257,6 +262,8 @@ trx_free(
 	ut_a(!trx->has_search_latch);
 	ut_a(!trx->auto_inc_lock);

+	ut_a(trx->dict_operation_lock_mode == 0);
+
 	if (trx->lock_heap) {
 		mem_heap_free(trx->lock_heap);
 	}
@ -1497,9 +1504,12 @@ trx_print(
  		default: buf += sprintf(buf, " state %lu", trx->conc_state);
  	}

+#ifdef UNIV_LINUX
+        buf += sprintf(buf, ", process no %lu", trx->mysql_process_no);
+#else
        buf += sprintf(buf, ", OS thread id %lu",
 		       os_thread_pf(trx->mysql_thread_id));
-
+#endif
 	if (ut_strlen(trx->op_info) > 0) {
 		buf += sprintf(buf, " %s", trx->op_info);
 	}
--- a/isam/pack_isam.c
+++ b/isam/pack_isam.c
@ -67,7 +67,7 @@ struct st_file_buffer {
  char *buffer,*pos,*end;
  my_off_t pos_in_file;
  int bits;
-  uint byte;
+  uint bytes;
 };

 struct st_huff_tree;
@ -1835,7 +1835,7 @@ static void init_file_buffer(File file, pbool read_buffer)
    file_buffer.pos=file_buffer.buffer;
    file_buffer.bits=BITS_SAVED;
  }
-  file_buffer.byte=0;
+  file_buffer.bytes=0;
 }


@ -1866,13 +1866,13 @@ static void write_bits (register ulong value, register uint bits)
 {
  if ((file_buffer.bits-=(int) bits) >= 0)
  {
-    file_buffer.byte|=value << file_buffer.bits;
+    file_buffer.bytes|=value << file_buffer.bits;
  }
  else
  {
    reg3 uint byte_buff;
    bits= (uint) -file_buffer.bits;
-    byte_buff=file_buffer.byte | (uint) (value >> bits);
+    byte_buff=file_buffer.bytes | (uint) (value >> bits);
 #if BITS_SAVED == 32
    *file_buffer.pos++= (byte) (byte_buff >> 24) ;
    *file_buffer.pos++= (byte) (byte_buff >> 16) ;
@ -1898,7 +1898,7 @@ static void write_bits (register ulong value, register uint bits)
    if (file_buffer.pos >= file_buffer.end)
      VOID(flush_buffer((uint) ~0));
    file_buffer.bits=(int) (BITS_SAVED - bits);
-    file_buffer.byte=(uint) (value << (BITS_SAVED - bits));
+    file_buffer.bytes=(uint) (value << (BITS_SAVED - bits));
  }
  return;
 }
@ -1910,7 +1910,7 @@ static void flush_bits (void)
  uint bits,byte_buff;

  bits=(file_buffer.bits) & ~7;
-  byte_buff = file_buffer.byte >> bits;
+  byte_buff = file_buffer.bytes >> bits;
  bits=BITS_SAVED - bits;
  while (bits > 0)
  {
@ -1918,7 +1918,7 @@ static void flush_bits (void)
    *file_buffer.pos++= (byte) (uchar) (byte_buff >> bits) ;
  }
  file_buffer.bits=BITS_SAVED;
-  file_buffer.byte=0;
+  file_buffer.bytes=0;
  return;
 }

--- a/Show More
+++ b/Show More