mysqld w/ --valgrind
- Fixed by implementing parsing of error log messages generated outside of
test case runs (eg. during server shutdown).
Also make mysql-test-run.pl not delete the error log after server restart,
which looses information about which warnings were found.
Finally, make multi_update2 a --big test.
mysql-test/lib/My/Test.pm:
Fix home-brewed (and broken) serialization in My::Test to use the standard
Storable serializer.
mysql-test/mysql-test-run.pl:
- Stop mysqld servers gracefully rather than kill -9 when
warnings are being checked.
- After stopping mysqld servers, do an additional parse of the error
log to check for any warnings generated during shutdown.
- Fix error log parsing to be careful not to skip parsing part of the
file, by keeping track of previous file position rather than
relying on mark_log markers.
- Workers report warnings during shutdown to the master process with
a new packet 'WARNINGS' which includes a list of names of test that
might have caused the problem (could be any test run since last
server start).
- Fail entire test suite if warnings are found.
- When we remove the server data dir before server restart, preserve the
error log (don't delete it between restarts), as it may contain
valuable information even for test cases which don't show direct
failures.
mysql-test/t/multi_update2.test:
Make test --big, as it takes a _long_ time to run and only tests a single bug.
The test itself is not faulty. The testcase timeout
problem happens if this IMHO mid size resource
(space in vardir, virtual memory, amount of disk I/O)
consuming test meets a weak (excessive disk I/O caused
by parallel applications or paging) testing box.
The modifications:
- Move the most time and disk I/O consuming subtest
for Bug 1820 into its own script (multi_update2)
This will reduce the likelihood that we exceed the
testcase timeout.
- Replace error numbers with error names
- Minor improvements of the formatting
-
