On Sat, Aug 15, 2020 at 12:37 AM Sérgio Basto <sergio(a)serjux.com> wrote:
On Fri, 2020-08-14 at 21:14 +0200, Fabio Valentini wrote:
> On Fri, Aug 14, 2020 at 8:48 PM Sérgio Basto <sergio(a)serjux.com>
> wrote:
> > On Fri, 2020-08-14 at 20:30 +0200, Fabio Valentini wrote:
> > > So, I've been trying to figure out strange build failures in my
> > > Java
> > > packages when running mock with `--enablerepo local`.
> > >
> > > First I thought java-11-openjdk update might be to blame, but now
> > > I
> > > found another "unrelated" issue:
> > >
> > > OCaml packages built with dune and `--enablerepo local` right now
> > > produce packages where binaries get permissions `555` set instead
> > > of
> > > `755` despite dune explicitly setting 755.
> > >
> > > So ... Huh?
> > >
> > > The only package update other than python 3.9 beta 5 → python 3.9
> > > rc
> > > 1
> > > and a Java update that happened in this time frame - which might
> > > affect *both* Java *and* OCaml builds in weird ways - would be
> > > ...
> > > the
> > > first glibc 2.33 development snapshot.
> >
> > It is against the new policy rules, please only update rawhide
> > after
> > completing the first branched compose and his mirror
> > synchronization. i.e when we have a new buildroot .
> >
> > While the compose today seems successful is not yet published
>
> Sérgio, I have no idea what you're talking about here. Do you mean
> the
> new policy to wait with a rawhide compose until the first "branched"
> compose is successful and synced out?
yes
> If I understand correctly, that
> "policy" is enforced by bodhi not allowing any builds into
"branched"
> by default, and releng not launching any rawhide composes until a
> "branched" compose finishes.
> I have no superpowers that allow me to circumvent those restrictions
> :)
> If I had, I would untag glibc 2.22.9000 to check if it is indeed
> broken, or if my development PC is haunted instead.
(snip)
https://bodhi.fedoraproject.org/updates/?packages=glibc
glibc-2.32.9000-1.fc34 has been submitted for stable by bodhi
a day ago ...
There's a difference between bodhi pushing a rawhide update to
"stable" and it being included in "rawhide" repositories.
Because bodhi does not compose rawhide. It just marks packages as "to
be included in the next compose".
In my view, if f33 has not yet been composed and published, all
packages that are in rawhide shouldn't have changes to what will be
f33, to allow for a smooth transaction and not make things more hard
that it is .
This is exactly what's already happening.
I think you're missing the "compose" step that's happening between
things getting built and them actually getting included in "public"
repositories.
Because that step has not happened for a few days, "to allow for a
smooth transaction", as you put it.
And I don't understand what is the rush of update things in
rawhide ,
why people can't wait one week until things calm down , is just an
exercise of coordination. And BTW, the opposite, we also should avoid
last minute changes, before mass-rebuild :)
I am in no rush, but I also don't want to wait for a week twiddling my
thumbs for no good reason.
PS: I agree, rushing changes into rawhide before the mass rebuild is a
bad idea, but that's a different problem ...
Hopefully f33 is almost published .
F33 won't be "published" for another two months :)
But if you're talking about a successful f33 compose, it already
happened half a day ago, and the repos and mirrors should get updated
soon ...
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.o...
Fabio