This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Distributions still suffering from s390 ABI change problems.


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 07/15/2014 12:35 PM, Siddhesh Poyarekar wrote:
> On Tue, Jul 15, 2014 at 11:48:33AM -0400, Carlos O'Donell wrote:
>> Fundamentally this is not our decision to make. Andreas Krebbel is the
>> machine maintainer and has the final say on when and how the ABI should
>> be changed for s390.
> 
> Agreed, although we as distribution maintainers would have to work
> with or around whatever is decided.

You do. You can also do whatever you want to support your users.

>> Given that the damage is done in 2.19, backing it out now is not a good
>> idea. We just have to move forward. I would not condone reverting the
>> changes given that 2.19 is already out the door.
> 
> Why not?  What is it about the change that is non-reversible?  2.19 is
> broken and nothing we do can reverse that.  That's no reason to not
> reverse the breakage if possible and introduce whatever feature is
> required in 2.21 with more thought given to ABI compatibility.

The 2.19 release is not broken, only certain configurations are broken.
If users recompiled their applications when their internal testing failed
then they have working systems using the new ABI. Changing this again
breaks those 2.19 based systems when they upgrade glibc. They already
went through the pain of fixing their systems, and now you break them
again. They are not going to be happy. We should aim for one break and
no more. The break happened.

>> We should:
>>
>> * Fix the pthread issues.
>> * Work to provide tooling warnings for future ABI issues.
> 
> The upstream-tracker ABI reports would have pointed this out, but they
> currently only do runs for x86_64 and i686[1].  Maybe we should look
> at extending that coverage somehow.

We should absolutely be running that for all RHEL/Fedora arches and using
Abigail to do ABI testing. We know we should head in that direction.

I think the result of some Cauldron 2014 talks will likely be more projects
around ABI testing :-)

Cheers,
Carlos.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJTxVorAAoJECXvCkNsKkr/oD8H/14NbVhOlcOVeeM8qp0g7fxK
E7ZwnVZpuhHowuFrkudjwNzfXWhkhv7+G8xtImMO+zG/vvWHi0gDcYjUNsryIpe8
6sCyN2JCsy3O/W5fe07fzuKntq9PTMvfcSwjQYure3JZDdM9dUNNtV+FEbS+awWt
xl++VeFVVmpBpWICX4KYH8E935e6i6vHmBK4mTuSMY3Xu6jeNsv4ZoS2coyWmpNb
lkAs5dHisUQBzCXkoQ/FxmA4V+0UrplQfZ+DCXgXDtvMswEKZyyd6muyvmTPGmXl
D5eWOPFsf35DTl/TgH2VOnVO1uG43JSSidWQU46c4y/AgCPokvI0wgm3JK/daHA=
=ga3t
-----END PGP SIGNATURE-----


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]