Re: Consensus summary around changing GLIBC PPC64 LE ABI default to 2.17

[Andreas Jaeger <aj at suse dot com>]

On 02/01/2014 12:57 AM, Carlos O'Donell wrote:
> The decision to switch the ABI default to 2.17 has been made by IBM here:
> https://sourceware.org/ml/libc-alpha/2014-01/msg00799.html
> 
> This email is to record final consensus around this issue.
> 
> ----------------------------------------------------------------------
> Consensus summary around changing GLIBC PPC64 LE ABI default to 2.17
> ----------------------------------------------------------------------
> 
> Please correct me if I'm wrong.
> 
> IBM (Steven Munroe) - Yay.
> 
> Red Hat (Carlos O'Donell) - Yay.
> 	- Sustained opposition to leaving ABI baseline at GLIBC_2.18.
> 	- Provided patches to help others rebuild during ABI change.
> 
> SUSE (Andreas Jaeger) - Nay.
> 	- Needs to rebuild distribution with new ABI.
> 
> Canonical (Adam Conrad) - Nay.
> 	- Needs to rebuild distribution with new ABI.
> 
> Joseph Myers - No sustained opposition
> 	- Considers use of 2.19 ABI the only sensible policy.
> 	- Considers symbol backports easy. Example given Nios II.
> 
> Roland McGrath - No sustained opposition.
> 	- Considers use of 2.19 ABI the only sensible policy.
> 
> H.J. Lu - No sustained opposition
> 	- Considers symbol backports easy. Example given x32.
> 
> Brooks Moses - No sustained opposition.
> 
> The goal of the discussions were to reach some consensus regarding 
> the PPC64 LE ABI changes to support 2.17-based distributions.
> 
> The three main technical answers to the question of "Which ABI?" are:
> 
> (a) Distros rebase glibc on 2.19 and rebuild with GLIBC_2.19.
> 
> - Red Hat is unable to do this given their constraints. The assumption
>   is that it's also out of the picture for SUSE and Canonical which 
>   have 2.18-based distributions.

Note that at openSUSE we're using the current glibc 2.19 HEAD, so no
need for us to backport to an older release, we just take what's committed.

So, while rebasing is costly - like any change of the ABI -  there is no
contraint for us.

> (b) Stay on 2.1[78] sources using GLIBC_2.1[89] default ABI and backport
>     all 2.1[89] symbols to produce a 2.1[78]-based release whose ABI
>     is identical to the GLIBC_2.1[89] ABI released upstream.
> 
> - Despite statements by several developers that this is easy, no major
>   enterprise distribution has been released with a glibc patched like this.
>   There may be additional risk.
> 
> - It is expected that neither SUSE nor Canonical want to try have a hybrid
>   symbol release either. Thus setting the ABI baseline to GLIBC_2.19
>   places them in the same position it places other 2.17-based distributions
>   with GLIBC_2.18 as the default ABI.


> (c) Move the ABI baseline to GLIBC_2.17.
> 
> - This is what Red Hat and IBM propose to support 2.17-based and newer
>   distributions in the PPC64 LE ecosystem.
> 
> After input from all parties the decision has been made by the machine
> maintainer IBM to choose (c).


Andreas
-- 
 Andreas Jaeger aj@{suse.com,opensuse.org} Twitter/Identica: jaegerandi
  SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GF: Jeff Hawn,Jennifer Guild,Felix Imendörffer,HRB16746 (AG Nürnberg)
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126