Problems with Python bindings in KF5

[Albert Astals Cid]

El dissabte, 15 d’abril de 2017, a les 16:10:24 CEST, Luca Beltrame va 
escriure:
> (CC'ed to the author of the bindings)
> 
> Hello,
> 
> There are a few problems with the KF5 Python bindings, which emerged when
> people (including but not limited to myself) wanted to package them.
> 
> First of all, they are untestable in anything that's not a Debian based
> distro, because they don't look for the right version of clang (unless
> you're on Ubuntu or Debian, of course). D5291, D5290 and D5289 solve this
> issue.
> 
> Second, some of the builds of these bindings (ki18n in particular) are half
> broken: they're prone to races, sometimes they build, and on the next try
> they don't.  Even make -j1 is sometimes not enough[1].
> 
> This problem was also posted to release-team[2].
> 
> In addition, there's no form of dependency tracking in CMakeLists.txt,
> meaning that if you build bindings that rely on other bindings being there
> (for their sip files) you might get confusing build errors.
> 
> But, you say, the CI is green, so they should be OK, right?
> 
> No, the CI  *is not testing them*. They're not getting built because of lack
> of dependencies. I filed a ticket but it's tricky to add them to the CI
> image due their reliance on clang and their requirement of clang >= 3.8.
> 
> So, we have code committed:
> 
> - that only builds on Debian/Ubuntu unless ECM is patched;
> - that is prone to build races;
> - that is not even tested in our own CI.
> 
> The conclusions are left as an exercise to the reader.

This last sentence is more confontrational that it needs to be, software will 
have bugs, let's work to fix them instead of throwing accusations and blame 
around :)

Cheers,
  Albert

> 
> [1] https://build.opensuse.org/build/
> home:luca_b:branches:KDE:Unstable:Frameworks/openSUSE_Factory/x86_64/ki18n/
> _log
> [2] https://marc.info/?t=149108102300001&r=1&w=2