Re: GCC broken in rawhide?

[Florian Weimer]

On 01/30/2018 06:59 PM, Charalampos Stratakis wrote:

Noting here that python2 segfaults now when using profile guided optimizations 
on x86_64 since the latest gcc update.


This is a bug in Python 2.7:

  https://bugzilla.redhat.com/show_bug.cgi?id=1540316

We're working on fixing it.

Thanks,
Florian
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Charalampos Stratakis]

- Original Message -
> From: "Vít Ondruch" <vondr...@redhat.com>
> To: devel@lists.fedoraproject.org
> Sent: Tuesday, January 30, 2018 1:14:43 PM
> Subject: Re: GCC broken in rawhide?
> 
> 
> 
> Dne 30.1.2018 v 11:16 Jakub Jelinek napsal(a):
> > On Tue, Jan 30, 2018 at 11:11:02AM +0100, Ralf Corsepius wrote:
> >> On 01/30/2018 10:00 AM, Florian Weimer wrote:
> >>> On 01/30/2018 09:54 AM, Ralf Corsepius wrote:
> >>>>> annobin.spec now uses:
> >>>>>
> >>>>>    %undefine _annotated_build
> >>>>>
> >>>>> so at least the circular dependency is no longer there.  You still
> >>>>> have to remember to rebuild it when a new version of GCC comes out
> >>>>> however.
> >>>> ... which apparently has just happened.
> >>> Yes, Fedora 28 will use GCC 8.
> >>>
> >>>> Consequences are affecting all released versions of Fedora, because
> >>>> it's impossible to apply bugfix updates to Fedora < rawhide, due to
> >>>> the GCC-chaos in rawhide.
> >>> Please provide more context when reporting issues, otherwise we have a
> >>> hard time helping you.
> >> Seeming hundreds of packages currently carry broken deps, dnf is
> >> malfunctioning etc.
> > That is what we have the mass rebuild scheduled for (AFAIK this week).
> 
> I'd say that when major GCC lands in Fedora, what happened yesterday
> [1], it would not hurt to send a note to fedora-devel.
> 
> Vít
> 
> 
> [1] https://koji.fedoraproject.org/koji/buildinfo?buildID=1021625
> ___
> devel mailing list -- devel@lists.fedoraproject.org
> To unsubscribe send an email to devel-le...@lists.fedoraproject.org
> 

Agreed.

Noting here that python2 segfaults now when using profile guided optimizations 
on x86_64 since the latest gcc update. And that is on top of the breakage 
caused by
the Sun RPC removal and the change to a new libnsl within glibc, the 
replacement of libcrypt with libxcrypt, the -z defs flag. That applies for 
python3 as well. 

And this is due to too many invasive changes at the same time.

koschei is good in notifying us on those failures, however when so many core 
components of the distro get rebased or alter functionality
at the same time it really adds unnecessary overhead onto packagers when trying 
to pinpoint the exact cause for failures and which package caused it.

-- 
Regards,

Charalampos Stratakis
Software Engineer
Python Maintenance Team, Red Hat
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Vít Ondruch]

Dne 30.1.2018 v 11:16 Jakub Jelinek napsal(a):
> On Tue, Jan 30, 2018 at 11:11:02AM +0100, Ralf Corsepius wrote:
>> On 01/30/2018 10:00 AM, Florian Weimer wrote:
>>> On 01/30/2018 09:54 AM, Ralf Corsepius wrote:
> annobin.spec now uses:
>
>    %undefine _annotated_build
>
> so at least the circular dependency is no longer there.  You still
> have to remember to rebuild it when a new version of GCC comes out
> however.
 ... which apparently has just happened.
>>> Yes, Fedora 28 will use GCC 8.
>>>
 Consequences are affecting all released versions of Fedora, because
 it's impossible to apply bugfix updates to Fedora < rawhide, due to
 the GCC-chaos in rawhide.
>>> Please provide more context when reporting issues, otherwise we have a
>>> hard time helping you.
>> Seeming hundreds of packages currently carry broken deps, dnf is
>> malfunctioning etc.
> That is what we have the mass rebuild scheduled for (AFAIK this week).

I'd say that when major GCC lands in Fedora, what happened yesterday
[1], it would not hurt to send a note to fedora-devel.

Vít


[1] https://koji.fedoraproject.org/koji/buildinfo?buildID=1021625
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Tom Hughes]

On 30/01/18 10:50, Ralf Corsepius wrote:

On 01/30/2018 11:16 AM, Jakub Jelinek wrote:

On Tue, Jan 30, 2018 at 11:11:02AM +0100, Ralf Corsepius wrote:

On 01/30/2018 10:00 AM, Florian Weimer wrote:

On 01/30/2018 09:54 AM, Ralf Corsepius wrote:

annobin.spec now uses:

    %undefine _annotated_build

so at least the circular dependency is no longer there.  You still
have to remember to rebuild it when a new version of GCC comes out
however.


... which apparently has just happened.


Yes, Fedora 28 will use GCC 8.


Consequences are affecting all released versions of Fedora, because
it's impossible to apply bugfix updates to Fedora < rawhide, due to
the GCC-chaos in rawhide.


Please provide more context when reporting issues, otherwise we have a
hard time helping you.

Seeming hundreds of packages currently carry broken deps, dnf is
malfunctioning etc.


That is what we have the mass rebuild scheduled for (AFAIK this week).

How comes, I am observing these breakages right now, in the official 
rawhide repos?


They are preventing me to bugfix-update released fedoras, because the 
corresponding rawhide build are failing.


Well it's quite simple. The new gcc has changed the soname for 
libgfortran which means that everything build using fortran needs

to be rebuilt.

What Jakub is saying is that the mass rebuild will fix that by 
rebuilding those packages.


Tom

--
Tom Hughes (t...@compton.nu)
http://compton.nu/
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Ralf Corsepius]

On 01/30/2018 11:16 AM, Jakub Jelinek wrote:

On Tue, Jan 30, 2018 at 11:11:02AM +0100, Ralf Corsepius wrote:

On 01/30/2018 10:00 AM, Florian Weimer wrote:

On 01/30/2018 09:54 AM, Ralf Corsepius wrote:

annobin.spec now uses:

    %undefine _annotated_build

so at least the circular dependency is no longer there.  You still
have to remember to rebuild it when a new version of GCC comes out
however.


... which apparently has just happened.


Yes, Fedora 28 will use GCC 8.


Consequences are affecting all released versions of Fedora, because
it's impossible to apply bugfix updates to Fedora < rawhide, due to
the GCC-chaos in rawhide.


Please provide more context when reporting issues, otherwise we have a
hard time helping you.

Seeming hundreds of packages currently carry broken deps, dnf is
malfunctioning etc.


That is what we have the mass rebuild scheduled for (AFAIK this week).

How comes, I am observing these breakages right now, in the official 
rawhide repos?


They are preventing me to bugfix-update released fedoras, because the 
corresponding rawhide build are failing.


Ralf
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Jakub Jelinek]

On Tue, Jan 30, 2018 at 11:11:02AM +0100, Ralf Corsepius wrote:
> On 01/30/2018 10:00 AM, Florian Weimer wrote:
> > On 01/30/2018 09:54 AM, Ralf Corsepius wrote:
> > > > annobin.spec now uses:
> > > > 
> > > >    %undefine _annotated_build
> > > > 
> > > > so at least the circular dependency is no longer there.  You still
> > > > have to remember to rebuild it when a new version of GCC comes out
> > > > however.
> > > 
> > > ... which apparently has just happened.
> > 
> > Yes, Fedora 28 will use GCC 8.
> > 
> > > Consequences are affecting all released versions of Fedora, because
> > > it's impossible to apply bugfix updates to Fedora < rawhide, due to
> > > the GCC-chaos in rawhide.
> > 
> > Please provide more context when reporting issues, otherwise we have a
> > hard time helping you.
> Seeming hundreds of packages currently carry broken deps, dnf is
> malfunctioning etc.

That is what we have the mass rebuild scheduled for (AFAIK this week).

Jakub
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Ralf Corsepius]

On 01/30/2018 10:00 AM, Florian Weimer wrote:

On 01/30/2018 09:54 AM, Ralf Corsepius wrote:

annobin.spec now uses:

   %undefine _annotated_build

so at least the circular dependency is no longer there.  You still
have to remember to rebuild it when a new version of GCC comes out
however.


... which apparently has just happened.


Yes, Fedora 28 will use GCC 8.

Consequences are affecting all released versions of Fedora, because 
it's impossible to apply bugfix updates to Fedora < rawhide, due to 
the GCC-chaos in rawhide.


Please provide more context when reporting issues, otherwise we have a 
hard time helping you.
Seeming hundreds of packages currently carry broken deps, dnf is 
malfunctioning etc.


Eg.

Error:
 Problem 1: cannot install both libgfortran-8.0.1-0.6.fc28.x86_64 and 
libgfortran-7.3.1-1.fc28.x86_64
  - package gcc-gfortran-8.0.1-0.6.fc28.x86_64 requires 
libgfortran.so.5()(64bit), but none of the providers can be installed
  - package blacs-mpich-devel-2.0.2-23.fc27.x86_64 requires 
libgfortran.so.4()(64bit), but none of the providers can be installed

  - cannot install the best candidate for the job
 Problem 2: gcc-gfortran-7.3.1-1.fc28.i686 has inferior architecture
  - package mpich-devel-3.2.1-2.fc28.x86_64 requires gcc-gfortran, but 
none of the providers can be installed
  - package gcc-gfortran-7.3.1-1.fc28.x86_64 requires gcc = 
7.3.1-1.fc28, but none of the providers can be installed
  - package gcc-7.3.1-1.fc28.i686 requires cpp = 7.3.1-1.fc28, but none 
of the providers can be installed
  - package gcc-7.3.1-1.fc28.x86_64 requires cpp = 7.3.1-1.fc28, but 
none of the providers can be installed
  - package libtool-2.4.6-21.fc28.x86_64 requires gcc(major) = 8, but 
none of the providers can be installed
  - package gcc-8.0.1-0.6.fc28.x86_64 requires cpp = 8.0.1-0.6.fc28, 
but none of the providers can be installed
  - cannot install both cpp-7.3.1-1.fc28.x86_64 and 
cpp-8.0.1-0.6.fc28.x86_64
  - cannot install both cpp-8.0.1-0.6.fc28.x86_64 and 
cpp-7.3.1-1.fc28.x86_64

  - cannot install the best candidate for the job
  - package gcc-gfortran-8.0.1-0.6.fc28.x86_64 requires 
libgfortran.so.5()(64bit), but none of the providers can be installed
  - cannot install both libgfortran-8.0.1-0.6.fc28.x86_64 and 
libgfortran-7.3.1-1.fc28.x86_64
  - package blacs-openmpi-devel-2.0.2-23.fc27.x86_64 requires 
libgfortran.so.4()(64bit), but none of the providers can be installed
 Problem 3: package scalapack-openmpi-devel-2.0.2-23.fc27.x86_64 
requires libscalapack.so.2()(64bit)(openmpi-x86_64), but none of the 
providers can be installed
  - package scalapack-openmpi-2.0.2-23.fc27.x86_64 requires 
libgfortran.so.4()(64bit), but none of the providers can be installed
  - cannot install both libgfortran-8.0.1-0.6.fc28.x86_64 and 
libgfortran-7.3.1-1.fc28.x86_64
  - package gcc-gfortran-8.0.1-0.6.fc28.x86_64 requires 
libgfortran.so.5()(64bit), but none of the providers can be installed
  - package openmpi-devel-2.1.1-5.fc28.x86_64 requires gcc-gfortran, 
but none of the providers can be installed
  - package gcc-gfortran-7.3.1-1.fc28.i686 requires gcc = 7.3.1-1.fc28, 
but none of the providers can be installed
  - package gcc-7.3.1-1.fc28.i686 requires cpp = 7.3.1-1.fc28, but none 
of the providers can be installed
  - package gcc-7.3.1-1.fc28.x86_64 requires cpp = 7.3.1-1.fc28, but 
none of the providers can be installed
  - package gcc-gfortran-7.3.1-1.fc28.x86_64 requires gcc = 
7.3.1-1.fc28, but none of the providers can be installed
  - package gcc-c++-8.0.1-0.6.fc28.x86_64 requires gcc = 
8.0.1-0.6.fc28, but none of the providers can be installed
  - package gcc-8.0.1-0.6.fc28.x86_64 requires cpp = 8.0.1-0.6.fc28, 
but none of the providers can be installed
  - cannot install both cpp-7.3.1-1.fc28.x86_64 and 
cpp-8.0.1-0.6.fc28.x86_64
  - cannot install both cpp-8.0.1-0.6.fc28.x86_64 and 
cpp-7.3.1-1.fc28.x86_64

  - cannot install the best candidate for the job
(try to add '--allowerasing' to command line to replace conflicting 
packages or '--skip-broken' to skip uninstallable packages)

___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Florian Weimer]

On 01/30/2018 09:54 AM, Ralf Corsepius wrote:

annobin.spec now uses:

   %undefine _annotated_build

so at least the circular dependency is no longer there.  You still
have to remember to rebuild it when a new version of GCC comes out
however.


... which apparently has just happened.


Yes, Fedora 28 will use GCC 8.

Consequences are affecting all released versions of Fedora, because it's 
impossible to apply bugfix updates to Fedora < rawhide, due to the 
GCC-chaos in rawhide.


Please provide more context when reporting issues, otherwise we have a 
hard time helping you.


Thanks,
Florian
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Ralf Corsepius]

On 01/29/2018 04:45 PM, Richard W.M. Jones wrote:

On Mon, Jan 29, 2018 at 03:55:53PM +0100, Florian Weimer wrote:

On 01/29/2018 03:43 PM, Kevin Kofler wrote:

Is https://fedoraproject.org/wiki/Changes/Annobin (no user-visible
improvements, only yet another global distrowide size increase) really worth
the circular dependency nightmare (rebuilding annobin requires GCC, but GCC
is configured to not work without annobin) or is it time to drop the feature
and enact the contingency plan?


Yes, it is required for meeting our security hardening objectives.

Ideally, annobin would be built from the GCC source package, but
since GCC needs more than twelve hours to build on armhfp, that is
not really an option while we are still adjusting the information
that annobin collects.


annobin.spec now uses:

   %undefine _annotated_build

so at least the circular dependency is no longer there.  You still
have to remember to rebuild it when a new version of GCC comes out
however.


... which apparently has just happened.

Consequences are affecting all released versions of Fedora, because it's 
impossible to apply bugfix updates to Fedora < rawhide, due to the 
GCC-chaos in rawhide.


Ralf

___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Richard W.M. Jones]

On Mon, Jan 29, 2018 at 03:55:53PM +0100, Florian Weimer wrote:
> On 01/29/2018 03:43 PM, Kevin Kofler wrote:
> >Is https://fedoraproject.org/wiki/Changes/Annobin (no user-visible
> >improvements, only yet another global distrowide size increase) really worth
> >the circular dependency nightmare (rebuilding annobin requires GCC, but GCC
> >is configured to not work without annobin) or is it time to drop the feature
> >and enact the contingency plan?
> 
> Yes, it is required for meeting our security hardening objectives.
> 
> Ideally, annobin would be built from the GCC source package, but
> since GCC needs more than twelve hours to build on armhfp, that is
> not really an option while we are still adjusting the information
> that annobin collects.

annobin.spec now uses:

  %undefine _annotated_build

so at least the circular dependency is no longer there.  You still
have to remember to rebuild it when a new version of GCC comes out
however.

RIch.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-builder quickly builds VMs from scratch
http://libguestfs.org/virt-builder.1.html
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Florian Weimer]

On 01/29/2018 03:43 PM, Kevin Kofler wrote:

Is https://fedoraproject.org/wiki/Changes/Annobin (no user-visible
improvements, only yet another global distrowide size increase) really worth
the circular dependency nightmare (rebuilding annobin requires GCC, but GCC
is configured to not work without annobin) or is it time to drop the feature
and enact the contingency plan?


Yes, it is required for meeting our security hardening objectives.

Ideally, annobin would be built from the GCC source package, but since 
GCC needs more than twelve hours to build on armhfp, that is not really 
an option while we are still adjusting the information that annobin 
collects.


Thanks,
Florian
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Kevin Kofler]

Is https://fedoraproject.org/wiki/Changes/Annobin (no user-visible 
improvements, only yet another global distrowide size increase) really worth 
the circular dependency nightmare (rebuilding annobin requires GCC, but GCC 
is configured to not work without annobin) or is it time to drop the feature 
and enact the contingency plan?

I would say the latter.

Kevin Kofler
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Jonathan Wakely]

On 27/01/18 22:31 +, Jonathan Wakely wrote:

On 27/01/18 09:39 +0100, Florian Weimer wrote:

On 01/27/2018 01:15 AM, Jonathan Wakely wrote:

On 26/01/18 20:11 +0100, Florian Weimer wrote:

On 01/26/2018 07:50 AM, Ralf Corsepius wrote:
ATM all rawhide builds are failing for me, because autoconf's 
tests for CC are failing.


Yes, we had an internal dependency issue involving annobin (not 
related to RPM dependencies), fixed by an annobin rebuild.  
Later, a defective build of annobin was tagged into rawhide 
which had a file conflict, but this has since been resolved as 
well.


Is annobin-3.2-2.fc28 the fixed version?


Ah!  We currently have:

 nevra|  name
+-
annobin-2.0-3.fc27.x86_64  | Fedora/27/x86_64
annobin-3.1-1.fc28.src | Fedora/rawhide/source
annobin-3.1-1.fc28.x86_64  | Fedora/rawhide/x86_64
annobin-3.1-3.fc28.aarch64 | Fedora/28-build/aarch64
annobin-3.1-3.fc28.i686| Fedora/28-build/i386
annobin-3.1-3.fc28.ppc64le | Fedora/28-build/ppc64le
annobin-3.1-3.fc28.s390x   | Fedora/28-build/s390x
annobin-3.1-3.fc28.x86_64  | Fedora/28-build/x86_64
(8 rows)

rawhide repositories are composes, the *-build repositories come 
directly from Koji.  (I haven't mirrored everything, which is why 
the list is incomplete.)


Koji isn't currently generating repositories, so annobin-3.2-2.fc28 
hasn't showed up.  However, annobin-3.1-3.fc28 is also fine 
(annobin-3.2-1.fc28 has this particularly bug), as long as you use 
it with GCC 7.3.  So if you build using Koji use the mysteriously 
named “local” repository in mock, you should be able to build 
packages.


And what about submitting builds to koji, not building in mock?

I'm trying to rebuild dozens of packages with the new Boost update,
in the f28-boost side tag.


OK, 3.2-2 is in the side tag now.
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Jonathan Wakely]

On 27/01/18 09:39 +0100, Florian Weimer wrote:

On 01/27/2018 01:15 AM, Jonathan Wakely wrote:

On 26/01/18 20:11 +0100, Florian Weimer wrote:

On 01/26/2018 07:50 AM, Ralf Corsepius wrote:
ATM all rawhide builds are failing for me, because autoconf's 
tests for CC are failing.


Yes, we had an internal dependency issue involving annobin (not 
related to RPM dependencies), fixed by an annobin rebuild.  Later, 
a defective build of annobin was tagged into rawhide which had a 
file conflict, but this has since been resolved as well.


Is annobin-3.2-2.fc28 the fixed version?


Ah!  We currently have:

  nevra|  name
+-
annobin-2.0-3.fc27.x86_64  | Fedora/27/x86_64
annobin-3.1-1.fc28.src | Fedora/rawhide/source
annobin-3.1-1.fc28.x86_64  | Fedora/rawhide/x86_64
annobin-3.1-3.fc28.aarch64 | Fedora/28-build/aarch64
annobin-3.1-3.fc28.i686| Fedora/28-build/i386
annobin-3.1-3.fc28.ppc64le | Fedora/28-build/ppc64le
annobin-3.1-3.fc28.s390x   | Fedora/28-build/s390x
annobin-3.1-3.fc28.x86_64  | Fedora/28-build/x86_64
(8 rows)

rawhide repositories are composes, the *-build repositories come 
directly from Koji.  (I haven't mirrored everything, which is why the 
list is incomplete.)


Koji isn't currently generating repositories, so annobin-3.2-2.fc28 
hasn't showed up.  However, annobin-3.1-3.fc28 is also fine 
(annobin-3.2-1.fc28 has this particularly bug), as long as you use it 
with GCC 7.3.  So if you build using Koji use the mysteriously named 
“local” repository in mock, you should be able to build packages.


And what about submitting builds to koji, not building in mock?

I'm trying to rebuild dozens of packages with the new Boost update,
in the f28-boost side tag.

___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Philip Kovacs]

Is Koschei eventually going to rebuild with the corrected gcc?  I see 
dependency changes after the last failed build, but Koschei hasn't rebuilt yet. 

On Saturday, January 27, 2018 4:01 AM, Peter Robinson 
 wrote:
 

  ATM all rawhide builds are failing for me, because autoconf's tests for
 CC are failing.
>>>
>>>
>>> Yes, we had an internal dependency issue involving annobin (not related
>>> to RPM dependencies), fixed by an annobin rebuild.  Later, a defective build
>>> of annobin was tagged into rawhide which had a file conflict, but this has
>>> since been resolved as well.
>>
>>
>> Is annobin-3.2-2.fc28 the fixed version?
>
>
> Ah!  We currently have:
>
>            nevra            |          name
> +-
>  annobin-2.0-3.fc27.x86_64  | Fedora/27/x86_64
>  annobin-3.1-1.fc28.src    | Fedora/rawhide/source
>  annobin-3.1-1.fc28.x86_64  | Fedora/rawhide/x86_64
>  annobin-3.1-3.fc28.aarch64 | Fedora/28-build/aarch64
>  annobin-3.1-3.fc28.i686    | Fedora/28-build/i386
>  annobin-3.1-3.fc28.ppc64le | Fedora/28-build/ppc64le
>  annobin-3.1-3.fc28.s390x  | Fedora/28-build/s390x
>  annobin-3.1-3.fc28.x86_64  | Fedora/28-build/x86_64
> (8 rows)
>
> rawhide repositories are composes, the *-build repositories come directly
> from Koji.  (I haven't mirrored everything, which is why the list is
> incomplete.)
>
> Koji isn't currently generating repositories, so annobin-3.2-2.fc28 hasn't
> showed up.  However, annobin-3.1-3.fc28 is also fine (annobin-3.2-1.fc28 has
> this particularly bug), as long as you use it with GCC 7.3.  So if you build
> using Koji use the mysteriously named “local” repository in mock, you should
> be able to build packages.

The newRepo issue is known by rel-eng/infra and is being investigated/worked on.

P
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org


   ___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Peter Robinson]

 ATM all rawhide builds are failing for me, because autoconf's tests for
 CC are failing.
>>>
>>>
>>> Yes, we had an internal dependency issue involving annobin (not related
>>> to RPM dependencies), fixed by an annobin rebuild.  Later, a defective build
>>> of annobin was tagged into rawhide which had a file conflict, but this has
>>> since been resolved as well.
>>
>>
>> Is annobin-3.2-2.fc28 the fixed version?
>
>
> Ah!  We currently have:
>
>nevra|  name
> +-
>  annobin-2.0-3.fc27.x86_64  | Fedora/27/x86_64
>  annobin-3.1-1.fc28.src | Fedora/rawhide/source
>  annobin-3.1-1.fc28.x86_64  | Fedora/rawhide/x86_64
>  annobin-3.1-3.fc28.aarch64 | Fedora/28-build/aarch64
>  annobin-3.1-3.fc28.i686| Fedora/28-build/i386
>  annobin-3.1-3.fc28.ppc64le | Fedora/28-build/ppc64le
>  annobin-3.1-3.fc28.s390x   | Fedora/28-build/s390x
>  annobin-3.1-3.fc28.x86_64  | Fedora/28-build/x86_64
> (8 rows)
>
> rawhide repositories are composes, the *-build repositories come directly
> from Koji.  (I haven't mirrored everything, which is why the list is
> incomplete.)
>
> Koji isn't currently generating repositories, so annobin-3.2-2.fc28 hasn't
> showed up.  However, annobin-3.1-3.fc28 is also fine (annobin-3.2-1.fc28 has
> this particularly bug), as long as you use it with GCC 7.3.  So if you build
> using Koji use the mysteriously named “local” repository in mock, you should
> be able to build packages.

The newRepo issue is known by rel-eng/infra and is being investigated/worked on.

P
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Florian Weimer]

On 01/27/2018 01:15 AM, Jonathan Wakely wrote:

On 26/01/18 20:11 +0100, Florian Weimer wrote:

On 01/26/2018 07:50 AM, Ralf Corsepius wrote:
ATM all rawhide builds are failing for me, because autoconf's tests 
for CC are failing.


Yes, we had an internal dependency issue involving annobin (not 
related to RPM dependencies), fixed by an annobin rebuild.  Later, a 
defective build of annobin was tagged into rawhide which had a file 
conflict, but this has since been resolved as well.


Is annobin-3.2-2.fc28 the fixed version?


Ah!  We currently have:

   nevra|  name
+-
 annobin-2.0-3.fc27.x86_64  | Fedora/27/x86_64
 annobin-3.1-1.fc28.src | Fedora/rawhide/source
 annobin-3.1-1.fc28.x86_64  | Fedora/rawhide/x86_64
 annobin-3.1-3.fc28.aarch64 | Fedora/28-build/aarch64
 annobin-3.1-3.fc28.i686| Fedora/28-build/i386
 annobin-3.1-3.fc28.ppc64le | Fedora/28-build/ppc64le
 annobin-3.1-3.fc28.s390x   | Fedora/28-build/s390x
 annobin-3.1-3.fc28.x86_64  | Fedora/28-build/x86_64
(8 rows)

rawhide repositories are composes, the *-build repositories come 
directly from Koji.  (I haven't mirrored everything, which is why the 
list is incomplete.)


Koji isn't currently generating repositories, so annobin-3.2-2.fc28 
hasn't showed up.  However, annobin-3.1-3.fc28 is also fine 
(annobin-3.2-1.fc28 has this particularly bug), as long as you use it 
with GCC 7.3.  So if you build using Koji use the mysteriously named 
“local” repository in mock, you should be able to build packages.


Thanks,
Florian
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Jonathan Wakely]

On 26/01/18 20:11 +0100, Florian Weimer wrote:

On 01/26/2018 07:50 AM, Ralf Corsepius wrote:
ATM all rawhide builds are failing for me, because autoconf's tests 
for CC are failing.


Yes, we had an internal dependency issue involving annobin (not 
related to RPM dependencies), fixed by an annobin rebuild.  Later, a 
defective build of annobin was tagged into rawhide which had a file 
conflict, but this has since been resolved as well.


Is annobin-3.2-2.fc28 the fixed version?

I've been waiting hours for it to appear in the repo but no luck, so
all my builds are failing with the file conflict.
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Florian Weimer]

On 01/26/2018 07:50 AM, Ralf Corsepius wrote:
ATM all rawhide builds are failing for me, because autoconf's tests for 
CC are failing.


Yes, we had an internal dependency issue involving annobin (not related 
to RPM dependencies), fixed by an annobin rebuild.  Later, a defective 
build of annobin was tagged into rawhide which had a file conflict, but 
this has since been resolved as well.


Sorry about that.

Thanks,
Florian
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Michael Schwendt]

On Fri, 26 Jan 2018 14:52:24 +0100, Remi Collet wrote:

> DEBUG util.py:439:  Error: Transaction check error:
> DEBUG util.py:439:file /usr/share/info/dir conflicts between
> attempted installs of annobin-3.2-1.fc28.x86_64 and info-6.5-1.fc28.x86_64
> 
> Seems related to
> https://src.fedoraproject.org/cgit/rpms/annobin.git/commit/?id=6b89f3290cbc9d2419bce617928b2f78b54836b8
> 

If the package would rebuild, I would have fixed it to fix the buildroot,
but it fails in the test suite.

diff --git a/annobin.spec b/annobin.spec
index f7891a6..1be0937 100644
--- a/annobin.spec
+++ b/annobin.spec
@@ -3,7 +3,7 @@
 Name:annobin
 Summary: Binary annotation plugin for GCC
 Version: 3.2
-Release: 1%{?dist}
+Release: 2%{?dist}
 
 License: GPLv3+
 URL: https://fedoraproject.org/wiki/Toolchain/Watermark
@@ -18,6 +18,9 @@ Source:  
https://nickc.fedorapeople.org/annobin-%{version}.tar.xz
 # This is a gcc plugin, hence gcc is required.
 Requires: gcc
 
+Requires(post): info
+Requires(preun): info
+
 BuildRequires: gcc-plugin-devel pkgconfig coreutils
 
 %description
@@ -63,12 +66,21 @@ touch doc/annobin.info
 
 %install
 %make_install
+rm -f %{_buildroot}%{_infodir}/dir
 
 %if %{with tests}
 %check
 make check
 %endif
 
+%post
+/sbin/install-info %{_infodir}/%{name}.info %{_infodir}/dir || :
+
+%preun
+if [ $1 = 0 ] ; then
+  /sbin/install-info --delete %{_infodir}/%{name}.info %{_infodir}/dir || :
+fi
+
 %files
 %{ANNOBIN_PLUGIN_DIR}
 %{_bindir}/built-by.sh
@@ -78,8 +90,7 @@ make check
 %exclude %{_datadir}/doc/annobin-plugin/COPYING3
 %exclude %{_datadir}/doc/annobin-plugin/LICENSE
 %doc %{_datadir}/doc/annobin-plugin/annotation.proposal.txt
-%{_infodir}
-%doc %{_infodir}/annobin.info.gz
+%{_infodir}/annobin.info*
 
 
#-
 
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Remi Collet]

Le 26/01/2018 à 12:42, Richard W.M. Jones a écrit :
> This build of annobin has apparently worked: 
> https://koji.fedoraproject.org/koji/taskinfo?taskID=24466259
> 
> So you can try building stuff which uses gcc now.

Some builds ok, but now

DEBUG util.py:439:  Error: Transaction check error:
DEBUG util.py:439:file /usr/share/info/dir conflicts between
attempted installs of annobin-3.2-1.fc28.x86_64 and info-6.5-1.fc28.x86_64

Seems related to
https://src.fedoraproject.org/cgit/rpms/annobin.git/commit/?id=6b89f3290cbc9d2419bce617928b2f78b54836b8

Remi

> ___
> devel mailing list -- devel@lists.fedoraproject.org
> To unsubscribe send an email to devel-le...@lists.fedoraproject.org
> 
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Richard W.M. Jones]

This build of annobin has apparently worked: 
https://koji.fedoraproject.org/koji/taskinfo?taskID=24466259

So you can try building stuff which uses gcc now.
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Richard W.M. Jones]

... which failed because GCC is broken in Rawhide.  Someone else with more 
powers is going to have to fix this I'm afraid.
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Richard W.M. Jones]

I looked at the annobin spec file and it doesn't appear to contain the version 
of GCC, so I went ahead and just issued a new build:

https://koji.fedoraproject.org/koji/taskinfo?taskID=24462980
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Ralf Corsepius]

On 01/26/2018 08:13 AM, Philip Kovacs wrote:

I'm getting:

configure: error: C compiler cannot create executables


This is exactly what I get.

Having a look into the corresponing config.log gives the error related 
annobin, I was referring to.


Ralf

___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Ralf Corsepius]

On 01/26/2018 08:05 AM, Samuel Sieb wrote:

On 01/25/2018 10:50 PM, Ralf Corsepius wrote:

Digging into details led me to this error:
...
cc1: error: fail to initialize plugin 
/usr/lib/gcc/x86_64-redhat-linux/7/plugin/annobin.so
annobin: conftest.c: Error: plugin built for compiler version (7.3.1) 
but run with compiler version (7.2.1)

...

AFAIS, GCC in rawhide was updated to GCC-7.3.1, but apparently annobin 
wasn't rebuilt, resulting into rawhide now carrying an inconsistent 
GCC-toolchain.


Are you sure you're reading that correctly?

Yes, I am.

# rpm -qa gcc* anno*
annobin-3.1-1.fc28.x86_64
gcc-c++-7.3.1-1.fc28.x86_64
gcc-gfortran-7.3.1-1.fc28.x86_64
gcc-7.3.1-1.fc28.x86_64

# grep annobin: config.log
annobin: conftest.c: Error: plugin built for compiler version (7.3.1) 
but run with compiler version (7.2.1)


  That message tells me that 
annobin was rebuilt for 7.3.1, but the gcc in the build environment is 
still 7.2.1.


cf. above.

Ralf
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Philip Kovacs]

I'm getting:
configure: error: C compiler cannot create executables 

On Friday, January 26, 2018 1:53 AM, Ralf Corsepius  
wrote:
 

 Hi,

ATM all rawhide builds are failing for me, because autoconf's tests for 
CC are failing.

Digging into details led me to this error:
...
cc1: error: fail to initialize plugin 
/usr/lib/gcc/x86_64-redhat-linux/7/plugin/annobin.so
annobin: conftest.c: Error: plugin built for compiler version (7.3.1) 
but run with compiler version (7.2.1)
...

AFAIS, GCC in rawhide was updated to GCC-7.3.1, but apparently annobin 
wasn't rebuilt, resulting into rawhide now carrying an inconsistent 
GCC-toolchain.

Ralf
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org


   ___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

Re: GCC broken in rawhide?

[Samuel Sieb]

On 01/25/2018 10:50 PM, Ralf Corsepius wrote:

Digging into details led me to this error:
...
cc1: error: fail to initialize plugin 
/usr/lib/gcc/x86_64-redhat-linux/7/plugin/annobin.so
annobin: conftest.c: Error: plugin built for compiler version (7.3.1) 
but run with compiler version (7.2.1)

...

AFAIS, GCC in rawhide was updated to GCC-7.3.1, but apparently annobin 
wasn't rebuilt, resulting into rawhide now carrying an inconsistent 
GCC-toolchain.


Are you sure you're reading that correctly?  That message tells me that 
annobin was rebuilt for 7.3.1, but the gcc in the build environment is 
still 7.2.1.

___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org

GCC broken in rawhide?

[Ralf Corsepius]

Hi,

ATM all rawhide builds are failing for me, because autoconf's tests for 
CC are failing.


Digging into details led me to this error:
...
cc1: error: fail to initialize plugin 
/usr/lib/gcc/x86_64-redhat-linux/7/plugin/annobin.so
annobin: conftest.c: Error: plugin built for compiler version (7.3.1) 
but run with compiler version (7.2.1)

...

AFAIS, GCC in rawhide was updated to GCC-7.3.1, but apparently annobin 
wasn't rebuilt, resulting into rawhide now carrying an inconsistent 
GCC-toolchain.


Ralf
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org