[ITP] cmocka 1.1.7 - C unit testing framework

[Jon Turney]

On 12/07/2024 03:31, Brian Inglis via Cygwin-apps wrote:
> On 2024-07-10 15:05, Jon Turney wrote:
>> On 10/07/2024 00:35, Brian Inglis via Cygwin-apps wrote:
>>> Description:
>>> Elegant unit testing framework for C with support for mock objects,
>>> derived from Google Cmockery.
>>>
>>> License:    Apache-2.0
>>>
>>> I would like to provide a Cygwin package for cmocka, as it is now
>>> required for testing my fortune-mod package.
>>>
>>> For more information see the project home pages:
>>>
>>>      https://cmocka.org
>>>
>>> It is packaged by major Linux and BSD distros:
>>>
>>>      https://repology.org/project/cmocka/versions
>>>
>>> There is an error building the docs using doxygen, and that hangs, using
>>> either cmake/make or cmake/ninja builds, locally and in GH CI (see
>>> previous job 8341 - had to cancel after 3 hours), with current stable 
>>> and a local build of the latest doxygen release, and an issue has 
>>> been raised upstream.
>>
>> I suspect this is a problem in cygwin.
>>
>> There is (what sounds like) a similar problem when building libxcb's 
>> documentation, in that it gets stuck somewhere in doxygen, waiting for 
>> a load of dot.exe processes to finish even after they're all done.
>>
>> I bisected that to working under 3.3.6 and failing under 3.4.0, but 
>> there's nothing obviously related in that interval, and I haven't 
>> found the time to bisect the commits...
> 
> Thanks Jon,
> 
> They agree with that suspicion in doxygen:
> 
>      https://github.com/doxygen/doxygen/issues/10251
> 
> Looking at some hints and approaches for doxygen diagnostics and debugging:
> 
>      -DDOXYGEN_DOT_NUM_THREADS=1
> 
> Eliminating all other job paramaters, that is the one that switches 
> hangs on and off.
> When you have time, would you care to retry libxcb docs with that 
> CYGCMAKE_ARGS or equivalent?
> I can see if upstream can help us narrow the code down to a simple STC 
> for us, or should we loop in, or punt to, Ken as maintainer? ;^>

I think the most productive investigation at this stage would be to 
confirm my observation of the range of commits in cygwin which 
introduces the regression, and try to bisect that

(Due to version skew, you'd probably have to hop in your time machine to 
make an installation using 3.3.4, then use that to build and test 
intermediate versions)