Re: gcc-defaults, gcc-3.3, gcc-3.4 for testing
Steve Langasek writes:
> Hi Matthias,
> On Sat, Mar 26, 2005 at 01:50:02PM +0100, Matthias Klose wrote:
> > gcc-3.3 and gcc-3.4 were addressed in
> > http://lists.debian.org/debian-release/2005/03/msg00027.html
> > the gcc-3.3 changes (compared to -8) include two m68k specific fixes
> > concerning wrong generation of code, and one hppa64 related update
> > (PR19697). more on the other noise in the update in a separate
> > message.
> Does "in a separate message" mean there's more information about these
> changes coming?
sorry the delay. the noise doesn't affect the transition of -12 to
testing. As Kurt pointed out in
some packages may have slipped into sarge, which were built with
gcc-3.3 3.3.5-9. AFAIK you are unable to tell from the stripped
binaries, that it's built by this compiler version. Maintainers may
still use this compiler package ... Some buildds did pick up this
compiler version as well, namely mips and mipsel packages were
uploaded. The s390 buildd was upgraded as well, but AFAIK packages
built with this gcc were not uploaded to the archive, but rebuilt a
second time (Gerhard, please confirm).
I don't have a good solution, maybe just file bug reports, severity
important, that the uploads for the packages listed by Kurt may be
wrongly built in the time period from 2004-03-04 (-9 upload) until
2005-03-13 (the last -10 upload, arm). We do not have to extend the
period until the s390 -10 upload, provided that the s390 packages were
rebuilt. The bug reports should ask for:
- checking, if the uploaded binaries were built using -9
- checking, if the buildd's built some binaries using -9. This task
could be done much easier by someone having access to the file
system with the build logs.
- if in doubt, rebuild/reupload the package.