Re: [pdf-devel] Build Error - unit test cases

[Aleksander Morgado]

Hi All,


> In fact this isn't the main issue in the mail I mentioned. I wasn't
> talking about the errors I got. I was talking about something Jose
> said there.
> In that mail  Jose says:
> 
>> The compilation of the unit tests fails when compiling
>> torture/unit/base/stm/pdf-stm-write.c, due to the usage of check
>> fixtures introduced by a recent patch.  Currently nocheck is not
>> supporting fixtures.
> 
> And then comes some talking about check and nocheck.
> Please read the mail again.
> Sorry if I wasn't clear enough.

Yes, that's all true. We developed builtin no-check as a very simple
replacement of `check' library so that it could be used in Windows
systems (or in systems where `check' was just not available). In fact,
if you check the source code of no-check, it's just a more-or-less API
compatible module replacing `check' library.


>>
>> Build Error
>> ---------------------
>>> gcc -DHAVE_CONFIG_H -I. -I../../src  -I../../lib
>> -I../../src -I../../src/base -I../../src/object
>> -I../../torture/tortutils -DTEST_DATA_PATH=\"../../torture/testdata\"
>> -Inocheck/  -g -g -MT pdf-stm-write.o -MD -MP -MF
>> .deps/pdf-stm-write.Tpo -c -o pdf-stm-write.o `test -f
>> 'base/stm/pdf-stm-write.c' || echo './'`base/stm/pdf-stm-write.c
>>> base/stm/pdf-stm-write.c: In function 'mem_stm_fixture_setup':
>>> base/stm/pdf-stm-write.c:52: error: 'test_name' undeclared (first use in 
>>> this function)
>>> base/stm/pdf-stm-write.c:52: error: (Each undeclared identifier is reported 
>>> only once
>>> base/stm/pdf-stm-write.c:52: error: for each function it appears in.)
>>> make[2]: ***

Some time ago, some new tests were added using fixtures in `check', and
fixtures were not originally added in our no-check replacement. And we
don't want to add support to them now (not sure if possible), because
now the original `check' library (at least the repository version they
have) works on Windows machines. This is, you can create your `check'
DLL (if not already created by someone out there) and use it along with
our tests in Windows OS. This new Windows-compatible check still lacks
forking support, so the behavior will be completely similar to what we
previously had with no-check, but supporting more things, like fixtures
(I hope so, I didn't try it).

Should we maybe disable `no-check' from now on? If it doesn't even
compile, it doesn't make much sense to have it there, I would say. Or
maybe we can hack it so that at least it ignores fixtures while compiling...


Cheers!
-Aleksander