[arch-dev-public] Doing a rebuild of [core]?
allan at archlinux.org
Thu Nov 18 07:18:39 CET 2010
On 18/11/10 15:52, Andreas Radke wrote:
> Am Thu, 18 Nov 2010 14:16:34 +1000
> schrieb "Allan McRae"<allan at archlinux.org>:
>> I was thinking of doing a fairly large rebuild of packages in [core]
>> for the following reasons:
>> - The toolchain is quite good at the moment and many packages have
>> not been built in a long time so could use a refresher build to take
>> advantage of what the newer toolchain offers (~15 packages are> 1
>> year old). I expect a toolchain update to happen in the next few
>> weeks so this is a good time.
> What update do you expect? Usually a rebuild is useful to catch
> up latest major gcc improvements. Gcc4.6 is still in an early
> stage. Maybe we should delay it until the gcc4.6 release.
I expect binutils in the next couple of weeks (they have branched) and
probably a glibc release given Fedora 14 is released with what is
usually considered an RC glibc release.
Gcc-4.6 is actually in stage 3 (bug fix only), but they still have a
large number of P1 bugs to get rid of before release. But given it took
a few months to sort out the toolchain issues after the initial gcc-4.5
release, I would much prefer doing this before gcc-4.6...
> The toolchain improvements over the last 2 years were of minor
> advantages. So I don't expect major performance improvements.
> Getting the packages into xz compression format and pkg pool would be
> nice though.
> One concern for a rebuild right now: most packagers allowed for
> pushing packages into core haven't been much online lately. Not sure
> if they can spend that much time. Sadly "makeworld" has died. But I
> remember Daniel made such a rebuild script locally (was at gcc4.3
Fair point. I was never expecting much help actually doing the rebuilds
as this is a very low priority task... and keeping [core] as pristine as
possible is probably not much of an obsession for people other than me!
Also, there is no real rush in getting this done. So I do not think
packaging manpower will be an issue as long as people can give signoffs
for the packages that are built.
More information about the arch-dev-public