Re: [PATCH] Add regression tests for abipkgdiff using ctf info

["Guillermo E. Martinez" <guillermo.e.martinez@oracle.com>]

On Monday, July 4, 2022 2:20:24 AM CDT Dodji Seketeli wrote:
> Hello Guillermo,
> 
> "Guillermo E. Martinez" <guillermo.e.martinez@oracle.com> a écrit:
> 
> >> > 	* tests/test-diff-pkg-ctf.cc: Testsuite for abipkgdiff --ctf.
> >> 
> >> Why come up with a something else but test-diff-pkg.cc just to test RPMs
> >> using abipkgdiff --ctf?
> > Actually that was my first approach, but looking inside of the acts to perform
> > the abipkgdiff DWARFs test, the steps could be changed in the feature
> > (new steps in `perform' member function applied to DWARFs not applicable to CTF),
> 
> Sorry, I am not sure to understand this.  What are the steps that would
> exist for DWARF that would not in CTF, and that would have to be handled
> at the testing level, rather than at the level of abipkgdiff/libabigail
> itself?  This is a honest question.  I can't think of any such case
> right now.  Do you see any?
hmm Ok,  for instance CTF support debug info embedded in binary files by default,
so no need to use `set_debug_packages_path' (I know, I added isl-debuginfo-0.16 ..
as unit test entry because nowadays eu-strip remove .ctf section by default) ,
for sure this option is handled in abipkgdiff/libabigail but DWARF and CFT tests
share  `set_debug_packages_path' this will make sense just DWARF test in the
future of course.

> > so it was my motivation to split the tests. But of course these are my
> > thoughts, and I will follow you advices because you have much more
> > experience. So what do you think, worth split the tests for DWARF and CTF?
> >> I mean, can't we just use test-diff-pkg.cc and fill in entries made of
> >> instances of struct InOutSpec with InOutSpec::prog_option filled in with
> >> a string prefixed by " --ctf " ?
> >> 
> >> Of course, those ctf-related entries would be guarded by #ifdef WITH_CTF
> >> or the like.
> >> 
> >> 
> >> Wouldn't that work and be simpler?
> > For the current implementation, absolutely. 
> 
> Then I'd rather keep the simpler way for now.  Unless you think there is
> an absolute need for the more complicated choice right now, of course.
Ok, I will do so.

Thanks for your advices!,
guillermo