Re: Reverting the s390 jmp_buf/ucontext_t ABI change

[Dan HorÃk <dan at danny dot cz>]

On Fri, 01 Aug 2014 10:58:09 +0200
Andreas Jaeger <aj@suse.com> wrote:

> On 08/01/2014 10:53 AM, Dan HorÃk wrote:
> > On Fri, 01 Aug 2014 09:16:17 +0200
> > Andreas Jaeger <aj@suse.com> wrote:
> > 
> >> On 08/01/2014 08:50 AM, Dan HorÃk wrote:
> >>> Hi,
> >>>
> >>> I'm the main person behind Fedora for s390x and I have some
> >>> objections regarding this change. We had to cope with the breakage
> >>> in 2.19 early this year, it took some time to realize what's wrong
> >>> and we found a solution how to make it work. After some
> >>> rebootstrapping work we have used the planned mass rebuild to
> >>> clean everything. Due some nasty gcc 4.9 bugs the rebuild on s390
> >>> finishes during these days. And you are telling we should do it
> >>> again? Sorry, but it is not possible, there is only one mass
> >>> rebuild planned in Fedora per release and now it's also too late
> >>> when we are approaching Fedora 21 Alpha deadline. Fortunately I
> >>> still have the notes for the rebootstrap, so it can be done again
> >>> (for Fedora 22), but not now.
> >>
> >> Dan, I think you want this change - otherwise applications build
> >> for previous Fedora releases, RHEL 6 or RHEL 7 will not work on
> >> newer
> > 
> > I think we could live with it for one Fedora release in the case the
> > change got reverted (like is happening now), or for all upcoming
> > releases if it the would bring advantages for the future
> 
> That's what I think - it brings advantage for the future.

I meant living with the broken ABI for some time, because the original
idea of the 2.19 ABI change was about bringing future extensibility of
the jmpbuf struct


		Dan