[xml] XPath Regression Fails on Win32 for libxml2-2.6.19

[Rush Manbert <rush manbert com>]

Hello,

I have built libxml2 version 2.6.19 on Windows XP using the method 
described in the win32 subdirectory. The library builds, but it fails if 
I run the tests by specifying the tests target to nmake.

After a lot of fooling around, I finally modified win32\Makefile.msvc so 
I could see what was happening.

I replaced lines 347 through 360 with the following:
   @echo ## XPath regression tests part 2
   @for %%I in ($(XML_SRCDIR)\test\XPath\docs\*.*) do @( \
       @echo I: %%I &&\
       for %%J in ($(XML_SRCDIR)\test\XPath\tests\%%~nxI*.*) do @( \
           @echo   J: %%J &&\
           if not exist $(XML_SRCDIR)\result\XPath\tests\%%~nxJ ( \
               $(BINDIR)\testXPath.exe -f -i %%I %%J > 
$(XML_SRCDIR)\result\XPath\tests\%%~nxJ &&\
               findstr /C:"MEMORY ALLOCATED" .memdump | findstr 
/C:"MEMORY ALLOCATED : 0" > nul \
           ) ELSE ( \
               @echo   ELSE: &&\
               @echo     testXPAth &&\
               $(BINDIR)\testXPAth.exe -f -i %%I %%J 2>&1 > 
result.%%~nxJ &&\
               copy .memdump .memdump.%%~nxJ &\
               @echo     findstr &&\
               findstr /C:"MEMORY ALLOCATED" .memdump | findstr 
/C:"MEMORY ALLOCATED : 0">null &&\
               @echo     fc &&\
               fc $(XML_SRCDIR)\result\XPath\tests\%%~nxJ result.%%~nxJ 
>null & \
               @echo IF ERRORLEVEL: &\
               IF ERRORLEVEL 1 (echo Error: %%I %%J & exit 1) & \
               del result.%%~nxJ \
           )\
       )\
   )

With these changes, here is the interesting part of the output:
## XPath regression tests
## XPath regression tests part 2
I: ..\test\XPath\docs\chapters
J: ..\test\XPath\tests\chaptersbase
ELSE:
testXPAth
       1 file(s) copied.
findstr
J: ..\test\XPath\tests\chaptersprefol
ELSE:
testXPAth
       1 file(s) copied.
findstr
I: ..\test\XPath\docs\id
J: ..\test\XPath\tests\idsimple
ELSE:
testXPAth
       1 file(s) copied.
findstr
I: ..\test\XPath\docs\simple
J: ..\test\XPath\tests\simpleabbr
ELSE:
testXPAth
       1 file(s) copied.
findstr
J: ..\test\XPath\tests\simplebase
ELSE:
testXPAth
       1 file(s) copied.
findstr
I: ..\test\XPath\docs\str
I: ..\test\XPath\docs\usr1
J: ..\test\XPath\tests\usr1check
ELSE:
testXPAth
       1 file(s) copied.
findstr
I: ..\test\XPath\docs\vid
J: ..\test\XPath\tests\vidbase
ELSE:
testXPAth
       1 file(s) copied.
findstr
NMAKE : fatal error U1077: 'for' : return code '0x1'
Stop.

Each of my .memdump.* files is empty, which appears to make findstr 
fail, because fc never gets run. I guess the error return from findstr 
at the end of the for loop must be what makes nmake quit, because the IF 
ERRORLVEL was never executed.

If I change the "&&" at the end of the findstr line to "&", so that the 
fc always gets run, the tests pass and I get this output:
## XPath regression tests
## XPath regression tests part 2
I: ..\test\XPath\docs\chapters
J: ..\test\XPath\tests\chaptersbase
ELSE:
testXPAth
       1 file(s) copied.
findstr
fc
IF ERRORLEVEL:
J: ..\test\XPath\tests\chaptersprefol
ELSE:
testXPAth
       1 file(s) copied.
findstr
fc
IF ERRORLEVEL:
I: ..\test\XPath\docs\id
J: ..\test\XPath\tests\idsimple
ELSE:
testXPAth
       1 file(s) copied.
findstr
fc
IF ERRORLEVEL:
I: ..\test\XPath\docs\simple
J: ..\test\XPath\tests\simpleabbr
ELSE:
testXPAth
       1 file(s) copied.
findstr
fc
IF ERRORLEVEL:
J: ..\test\XPath\tests\simplebase
ELSE:
testXPAth
       1 file(s) copied.
findstr
fc
IF ERRORLEVEL:
I: ..\test\XPath\docs\str
I: ..\test\XPath\docs\usr1
J: ..\test\XPath\tests\usr1check
ELSE:
testXPAth
       1 file(s) copied.
findstr
fc
IF ERRORLEVEL:
I: ..\test\XPath\docs\vid
J: ..\test\XPath\tests\vidbase
ELSE:
testXPAth
       1 file(s) copied.
findstr
fc
IF ERRORLEVEL:


After making this change, I went and modified one of the files in 
..\test\XPath\tests and verified that the test fails as expected.

My conclusion is that if I just make the change from "&&" to "&" on the 
findstr command line, then things will work, but I really don't 
understand what the intent is for the findstr command. If I make the 
change, then we will discard the result of the findstr.

I didn't see anything in the archives about this, which makes me a 
little nervous. Has anyone else seen this under Windows and can anyone 
tell me whether the change I have made is the correct fix?

Thanks,
Rush