Le mercredi, 28 septembre 2011 23.48:47, Thomas Gaugler a écrit :
>
> > What do you think ?
>
> On one hand we could drop the parallel building support but on the
> other hand we would limit those who try out massive parallel building.
Those who try out can patch the parallel building support back in (We could
only "comment" it out instead of dropping it.
> An even simpler solution would be by not passing on
> DEB_BUILD_OPTIONS=parallel= from the caller program, e. g. omitting
> -j from dpkg-buildpackage.
Well; that's not possible. Each builddaemon uses its own configuration for
parallel building. What the Debian Policy mandates is that if parallel=n is
recognised, then parallel building must succeed.
> Furthermore I am not even sure if the issues are actually caused by
> the parallel building. I also experienced failures during the test stage
> in case of non-parallel building, although these failures only occurred
> very rarely. As far as I recall a mismatch of the calling
> convention (stdcall) was identified by me as likely candidate for the
> failures. I would need to dig out the change from the experimental
> unicode patch pushed into the git repository.
Ah, if you have an idea to solve this, I'm all ears !
> So I tend more to keep the parallel building support. However I am
> undecided.
Well, I try to keep in mind that we are preparing the next Debian stable
release and that packages that FTBFS are a problem, for licence and security
reasons, so I think we are best with slowly-non-parallel-built packages than
with fast-built-only-if-moon-is-aligned packages, hence the proposal to drop
the parallel building.
Cheers,
--
OdyX
signature.asc
Description: This is a digitally signed message part.