Bug#676906: nmu: atlas_3.8.4-6

2012-06-11 Thread Sébastien Villemot
Julien Cristau  writes:

> On Sun, Jun 10, 2012 at 19:07:38 +0200, Sébastien Villemot wrote:
>
>> What is clear is that the atlas binaries built on the amd64 buildds
>> should not have the problem that is affecting the current amd64 binary
>> (for example I tested the binary for 3.8.4-3 which was built on buildds
>> and it is sane). So a binNMU seems a good short-term solution.
>> 
> That's not clear to me.  We have different buildds with different CPUs,
> just because one version wasn't affected doesn't mean the next one won't
> be.

Looking again at atlas's build system, I realize that I was wrong: it
*is* selecting the more generic set of processor instructions. So there
is no need to modify the build system AFAIUI, sorry for the noise.

But then I don't understand why the package built by the maintainer is
broken.

-- 
Sébastien Villemot
Researcher in Economics & Debian Maintainer
http://www.dynare.org/sebastien
Phone: +33-1-40-77-84-04 - GPG Key: 4096R/381A7594


pgp2rySiERGrm.pgp
Description: PGP signature


Bug#676906: nmu: atlas_3.8.4-6

2012-06-11 Thread Julien Cristau
On Sun, Jun 10, 2012 at 19:07:38 +0200, Sébastien Villemot wrote:

> What is clear is that the atlas binaries built on the amd64 buildds
> should not have the problem that is affecting the current amd64 binary
> (for example I tested the binary for 3.8.4-3 which was built on buildds
> and it is sane). So a binNMU seems a good short-term solution.
> 
That's not clear to me.  We have different buildds with different CPUs,
just because one version wasn't affected doesn't mean the next one won't
be.

Cheers,
Julien



--
To UNSUBSCRIBE, email to debian-release-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20120611173114.gw31...@radis.cristau.org



Bug#676906: nmu: atlas_3.8.4-6

2012-06-10 Thread Sébastien Villemot
Philipp Kern  writes:

> On Sun, Jun 10, 2012 at 02:34:56PM +0200, Sébastien Villemot wrote:
>> The package for amd64 has apparently been built with a too recent and 
>> specific
>> CPU architecture, leading to "Illegal instructions" on many recent Intel Core
>> CPUs. A rebuild fixes the problem.
>
> How? By assuming that the buildd has a less recent CPU? That would be broken.
> (I.e. please fix the build system instead.) If the maintainer, however, had 
> put
> local overrides in place and built with them, then this request is valid (and
> the maintainer should be LARTed appropriately).

First, note that the decision to request a binNMU has been made jointly
with the maintainer (whom I am going to help to maintain this package).

What is clear is that the atlas binaries built on the amd64 buildds
should not have the problem that is affecting the current amd64 binary
(for example I tested the binary for 3.8.4-3 which was built on buildds
and it is sane). So a binNMU seems a good short-term solution.

But I agree with you that the long-term solution is to have a package
whose behaviour does not depend on the machine on which it was built. I
am currently discussing with Sylvestre how to achieve this.

Best,

-- 
Sébastien Villemot
Researcher in Economics & Debian Maintainer
http://www.dynare.org/sebastien
Phone: +33-1-40-77-84-04 - GPG Key: 4096R/381A7594


pgpFKPFfWVpjc.pgp
Description: PGP signature


Bug#676906: nmu: atlas_3.8.4-6

2012-06-10 Thread Philipp Kern
On Sun, Jun 10, 2012 at 02:34:56PM +0200, Sébastien Villemot wrote:
> The package for amd64 has apparently been built with a too recent and specific
> CPU architecture, leading to "Illegal instructions" on many recent Intel Core
> CPUs. A rebuild fixes the problem.

How? By assuming that the buildd has a less recent CPU? That would be broken.
(I.e. please fix the build system instead.) If the maintainer, however, had put
local overrides in place and built with them, then this request is valid (and
the maintainer should be LARTed appropriately).

Kind regards
Philipp Kern



-- 
To UNSUBSCRIBE, email to debian-release-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20120610142343.ga32...@hub.kern.lc



Bug#676906: nmu: atlas_3.8.4-6

2012-06-10 Thread Sébastien Villemot
Package: release.debian.org
Severity: normal
User: release.debian@packages.debian.org
Usertags: binnmu

The package for amd64 has apparently been built with a too recent and specific
CPU architecture, leading to "Illegal instructions" on many recent Intel Core
CPUs. A rebuild fixes the problem.

nmu atlas_3.8.4-6 . amd64 . -m "Rebuild with a more generic x86_64 CPU 
architecture (Closes: #676885)"

Thanks,

-- System Information:
Debian Release: wheezy/sid
  APT prefers testing
  APT policy: (990, 'testing'), (600, 'unstable'), (550, 'experimental')
Architecture: amd64 (x86_64)

Kernel: Linux 3.2.0-2-amd64 (SMP w/4 CPU cores)
Locale: LANG=fr_FR.UTF-8, LC_CTYPE=fr_FR.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash



-- 
To UNSUBSCRIBE, email to debian-release-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/20120610123456.15585.73293.report...@brouzouf.villemot.name