At least when merely public API is to be leveraged, one cannot use
xmlBufCreate function that would otherwise be a clear fit, and relying
on some invariants wrt. how some other struct fields will get
initialized along the construction/filling such parent struct and
(ab)using that instead does not appear clever, either.
Hence, instruct people what's the Right Thing for the moment, that is,
make them use xmlNodeDumpOutput instead (together with likewise public
xmlAllocOutputBuffer).
Going forward, it's questionable what do with xmlBuf* family of
functions that are once public, since they, for any practical purpose,
cannot be used by the library clients (that's how I've run into this).
Signed-off-by: Jan Pokorný <jpokorny@redhat.com>
This is a partial revert of commit c71f9305. I'm not sure what issue
this commit was trying to solve but it seems to be related to a
circular dependency. It might be related to tree.h being included
from dict.h which is unnecessary.
Resolves!22.
Make sure that all parameters and return values of hash callback
functions exactly match the callback function type. This is required
to pass clang's Control Flow Integrity checks and to allow compilation
to asm.js with Emscripten.
Fixes bug 784861.
In my previous commit that silenced some test output I didn't realize
that doc/examples/Makefile.am was autogenerated.
Also make index.py output deterministic by sorting the glob results.
- Silence test output.
- Clean up after doc/examples tests.
- Adjust expected output for script tests.
- Add missing results for relaxng/pattern3
There are still two test failures I can't comment on:
- regexp/bug316338
- schemas/any4_0
configure.am:
* Explicitly disallow --enable-rebuild-docs when builddir != srcdir, per
what you said about needing to build docs with an in-source build
doc/Makefile.am:
* Ensure that xmlversion.h is in the source tree before running
apibuild.py, to avoid generating an incomplete libxml2-api.xml
* Update the .PHONY target (forgot to do this earlier)
doc/devhelp/Makefile.am:
* Wrap the doc-generating rule in an "if REBUILD_DOCS" conditional so it
doesn't cause trouble for regular users
* Added a handy-dandy "rebuild" target
doc/examples/index.py:
* NOTE: You need to run this script to regenerate the files it creates,
and then commit the newly-updated files! The generated files currently
in git master (e.g. doc/examples/Makefile.am) are out of date even
before this patch!
* index.html really needs to be in EXTRA_DIST
* Wrap the doc-generating rules in an "if REBUILD_DOCS" conditional,
because they shouldn't be active otherwise
I use libxml xpath engine on quite large (and mostly "flat") xml files.
It seems that Shellsort, that is used in xmlXPathNodeSetSort is a
performance bottleneck for my case. I have read some posts about sorting
in libxml in the libxml archive, but I agree that qsort was not the way
to go. I experimented with Timsort instead and my results were good for
me. For about 10000 nodes, my test was about 5x faster with Timsort,
for 1000 nodes about 10% faster, for small data files, the difference
was not measurable.
* timsort.h: the algorithm, kept in a separate header
* xpath.c: plug in the new algorithm in xmlXPathNodeSetSort
* Makefile.am: add the header to the EXTRA_DIST
* doc/apibuild.py: avoid indexing the new header
As suggested by Andrew W. Nosenko:
Proposal: expose the new xmlBufShrink() to the "public" API for
compatibility with xmlBufUse().
Reason: the following scenario:
1. Read something into xmlParserInputBuffer (e.g. using
xmlParserInputBufferRead())
2. Extract content through xmlBufContent()
3. Extract content length through xmlBufUse(). Result have type
'size_t'.
4. Use this content
5. Now, you need to shrink the buffer. How to do it? Doing that
through legacy xmlBufferShrink() is unsafe because it uses 'unsigned
int' and the whole point of introducing the new API was handling the
cases, when 'unsigned int' is not enough. Therefore, need to use the
new xmlBufShrink(). But it is "private".
Therefore, I propose to expose the new xmlBufShrink() in the same way,
as xmlBufContent() and xmlBufUse() are exposed.
doc/Makefile.am:
* Build what's in doc/ before doc/devhelp/, as the dependency graph flows
that way
* Add "--path $(srcdir)" so that xsltproc can find DTDs in srcdir
* Replaced $(top_srcdir)/doc with an equivalent $(srcdir)
* Qualified libxml2-api.xml with $(srcdir) as it's always generated there
* Rewrote the dependencies for libxml2-api.xml so that xmlversion.h
doesn't throw everything off
doc/devhelp/Makefile.am:
* Use Automake constructs to install the HTML files instead of an
install-data-local rule
* Reorganized the file a bit (hello whitespace!)
* EXTRA_DIST doesn't need to list so many files now that dist_devhelp_DATA
is being used
* Only print "Rebuilding devhelp files" if rebuilding is actually
occurring
doc/examples/index.py:
* Make the "this file is auto-generated" banner more prominent
* Autotools updates: Use AM_CPPFLAGS/AM_CFLAGS instead of INCLUDES
* Got rid of DEPS as it's not needed (Automake already sees the dependency
on libxml2.la by way of LDADD(S))
* Replaced LDADDS with LDADD, which is applied to all programs listed
in the file. Since all the test programs have the same link
dependencies, this way is more concise yet equivalent.
* Remove the *.tmp files via "make clean" instead of having the test
programs do it themselves (more on this later)
* Invoke index.py in srcdir, as it pretty much needs to run there
* Restructured the index.html rule so that only the xmllint invocation is
allowed to fail
* Use $(MKDIR_P) instead of $(mkinstalldirs), $(VAR) instead of @VAR@
* Remove symlinks for test?.xml in an out-of-source build
* Sort lists for neatness
* Better formatting for EXTRA_DIST and noinst_PROGRAMS variables
* Simplified the Automake bits printed for each program: *_LDFLAGS doesn't
need to be specified as it's empty anyway, *_DEPENDENCIES is redundant,
*_LDADD isn't needed due to the global LDADD
* Added a bit that symlinks in test?.xml from srcdir in out-of-source
builds. This allows the reader4 test to read these files in the current
directory, which ensures that the output always looks the same (i.e.
does not contain references to srcdir)
* Don't hide the test program invocation (or else it's hard to tell which
test failed), and don't use superfluous parentheses
* NOTE: If you check in these changes, be sure to run this script and also
check in the updated files that it generates!
doc/examples/*.c:
* Updated the test: lines so that
+ "&&" is used to separate commands instead of ";" so that errors are
not masked
+ reference files are qualified with $(srcdir)/
+ no "rm" takes place -- these are a problem because (1) if a test
fails, it's useful to have the output file ready for inspection; (2)
the "rm" invocation masks a potential non-zero exit status from diff
(This is why I added the CLEANFILES line above)
doc/examples/io1.res:
* Updated this ref file so that the test passes. (This is correct, right?)
doc/examples/reader4.res:
* Changed this back to its original form, as the symlinking of test?.xml
means this file no longer has to contain path prefixes on the filenames
doc/examples/testWriter.c:
* Changed the output filenames to *.tmp instead of *.res, partly for
consistency, partly to not have to add special cases to CLEANFILES
doc/examples/xpath1.c:
* Removed the "./" prefix on the test invocation, which is redundant as
index.py already adds one
