Re: BBv2 tests failing... since when?

From: David Abrahams (dave_at_[hidden])
Date: 2005-01-21 11:47:39

Vladimir Prus <ghost_at_[hidden]> writes:

> On Tuesday 04 January 2005 17:52, David Abrahams wrote:
>
>> >> expected -------- all changes caused by last build command ----------
>> >> Added files : ..... 'bin/gcc-3.2/debug/a.exe',
>> >> 'bin/gcc-3.2/debug/a.obj']
>> >
>> > Try running
>> >
>> > python test_all.py gcc-3.2
>> >
>> > The test system does not know the default version of gcc is 3.2
>>
>> Doesn't work.
>>
>
>> generators_test : File bin/gcc-3.2/debug/a.obj not added as
>> expected
>> -------- all changes caused by last build command ----------
>> Added files : [..........
> .......
>> 'bin/gcc-3.2/debug/a.o', 
>
> Is this some cygwin system? 

No. At least, I'm launching the tests through an ordinary Windows NT
version of Python. I do have Cygwin installed, and many cygwin
components are in the path. My GCC is a cygwin GCC. If these tests
only work when "gcc" invokes a mingw GCC it will be very
inconvenient. It used to work just fine.

Incidentally, doing

using gcc : : c:/tools/mingw-2.0/bin/gcc.exe ;

yields

generators_test : "['c:\\boost\\tools\\build\\v2\\test\\../../jam_src\\bin.ntx86\\bjam', '-sBOOST_BUILD_PATH=c:\\boost\\tools\\build\\v2\\test', '-d0', '--quiet', 'gcc']" returned 1
STDOUT ============
"WD consuming" d.wd 
"Dolphin consuming" bin\gcc\debug\d_lexer.dlp 
"Whale consuming " bin\gcc\debug\d_parser.whl 
"X_PRO: source is " y.x_pro 
"X: source is " bin\gcc\debug\y.x1 bin\gcc\debug\y.x2 
...skipped <pbin\gcc\debug>a.exe for lack of <pbin\gcc\debug>x.o...
warning: using independent target <pbin\gcc\debug>e.positions
"CPP-MARK consuming " e.cpp 
"NM target source consuming " bin\gcc\debug\e.marked.cpp 
...failed updating 2 targets...

STDERR ============
bin/gcc/debug/x.c:1: parse error before string constant
bin\gcc\debug\a.o: In function `main':
c:/docume~1/dave/locals~1/temp/tmpb8ndms/a.cpp:16: undefined reference to `__gxx_personality_v0'

END ===============
FAILED
project_test3 : File bin/gcc/debug/a.obj not added as expected

I have been trying to check in some fixes for several weeks, but I
can't even establish a baseline that the current CVS codebase works.

> There's mismatch between produced extension ".o" and expected ".obj"
> -- which is computed in BoostBuild.py (function
> prepare_suffix_map). I'm sure this logic worked OK on pure NT, but I
> never succeeded in installing Cygwin.

I will try running the tests through Cygwin to see what happens.
Doesn't work there either:

core_import_module : PASSED
generators_test : "['c:\\boost\\tools\\build\\v2\\test\\../../jam_src\\bin.ntx86\\bjam', '-sBOOST_BUILD_PATH=c:\\boost\\tools\\build\\v2\\test', '-d0', '--quiet', 'gcc']" returned 1
STDOUT ============
"WD consuming" d.wd 
"Dolphin consuming" bin\gcc\debug\d_lexer.dlp 
"Whale consuming " bin\gcc\debug\d_parser.whl 
"X_PRO: source is " y.x_pro 
"X: source is " bin\gcc\debug\y.x1 bin\gcc\debug\y.x2 
...skipped <pbin\gcc\debug>a.exe for lack of <pbin\gcc\debug>x.o...
warning: using independent target <pbin\gcc\debug>e.positions
"CPP-MARK consuming " e.cpp 
"NM target source consuming " bin\gcc\debug\e.marked.cpp 
...failed updating 1 target...

STDERR ============
bin/gcc/debug/x.c:1: error: parse error before string constant

END ===============

-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com