Okay, I fixed the place in xpath.c where this broke predicate selection. Now I'm also wondering if there is other places in the
XPAth
code where this may show up, maybe something like 0 = NaN should be tested for this bug on the Microsoft platform too. I think those are part of the "make tests" regression tests but not run usually on the MSC or Borland compilers.
The attached patch to win32\Makefile.msvc starts the framework for performing regression tests on Win32/MSVC. Running nmake tests will run all the available regression tests, currenly only XPathtests and performs it as closely as possible to the unix tests, just using windows scriping instead of unix [ba]sh. I plan to add the other tests as well, but this is slow going as I have never written a single shell script before. Anyways, running nmake XPathtests with the current cvs head on Win32/msvc6 passes all tests. I'll post more patches as I add the various units. ?p
Attachment:
makefile.diff
Description: Binary data