Re: [gentoo-dev] Adding slot and subslot deps to others' packages

[Michał Górny]

Dnia 2014-03-02, o godz. 19:09:09
Jeroen Roovers <jer@g.o> napisał(a):

> On Sun, 2 Mar 2014 12:32:05 -0500
> Mike Gilbert <floppym@g.o> wrote:
>
> > On Sun, Mar 2, 2014 at 10:45 AM, Jeroen Roovers <jer@g.o>
> > wrote:
> > > On Sun, 2 Mar 2014 09:37:22 +0100
> > > Michał Górny <mgorny@g.o> wrote:
> > >
> > >> Few months ago I have written a small FAQ on how to use slots
> > >> and subslots for library dependencies properly [1]. However, today
> > >> I see that most of the developers didn't care to properly update
> > >> their packages and when I introduced binary compatibility slot in
> > >> libgcrypt, I had my hands full of work fixing the mess for a
> > >> single package.
> > >
> > > How about you file a tracker bug report for each library package,
> > > and then file bug reports per package using that dependency
> > > blocking the tracker bug?
> > >
> >
> > That is certainly the conservative way to handle this, and it seems
> > like a lot more work.
>
> Not managing the migration is apparently a lot of work, since the work
> isn't being done "voluntarily". Managing the migration might seem like a
> lot of work, but at least it allows you to track progress properly and
> tick off finished tasks, while in the mean time everyone involved gets
> informed, and perhaps gets their bugs fixed for them quicker.

Except that with that number of bugs, some of them are going to get
outdated very quickly due to version bumps where maintainer ignored
the bug.

And believe me, if I'll file around 50 bugs for each maintainer, they
are not going to be happy. I guess some of them will be closed as
DONTCARE, and others reassigned to me like I needed that. Pretty much
like you keep assigning every bug with 'clang' in summary line to me
like I was supposed to patch every package in the world for clang
compatibility.

> > If there are any reasonable objections (besides maintainer
> > territorial-ism), of course the QA team should consider them.
>
> That's back to the "agreement" bit, I guess.
>
> Honestly, setting up a tracker and blocking it with bugs about packages
> which someones-sub-SLOT-checking-script has vetted to be involved could
> be done in less than a day (for the hundred or so packages that depend
> on dev-libs/libgcrypt). It doesn't need some QA team to study in depth
> -- it needs a couple of volunteers to do the checks and file the bug
> reports.

I'm not talking about libgcrypt. Those dependencies were 'mostly fixed'
already and no sane person will revbump every single package just to
make sure that everything will go right. Especially when Council
banned a few EAPIs and the revbump would have to be connected with EAPI
bump... and that would really make it all so happy and awesome.

So I'm not talking about few dozen libgcrypt consumers. I'm talking
about all packages in Gentoo that depend on at least a single library.
I won't give you numbers but I guess you can imagine the number of
bugs.

-- 
Best regards,
Michał Górny