[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: gcc 4.1 or not

Martin Michlmayr writes:
> * Andreas Barth <aba@not.so.argh.org> [2006-06-04 21:01]:
> > As we are below the 20 packages count if bug #366820 is correct (and
> > Martin just confirmed the number), it is ok to do the switch now.
> > Martin, can you please also mark these bugs as serious now (as
> > they're FTBFS then)?
> Yes, we have been under 20 bugs for a while now.  Most of the
> outstanding bugs fall into two categories:
>  - bugs that have been fixed upstream and are now waiting for a new
>    version
>  - packages that have other RC bugs and cannot be uploaded
> Fortunately, the former is the majority.  All maintainers of these
> packages are aware of the importance of these bugs and I'll continue
> to pester them. ;-)
> While I've been building the archive with gcc 4.2 recently, I've
> haven't done a re-run with 4.1.1 yet.  Matthias uploaded the packages
> a few days ago and they passed NEW today so I'll do it this week.
> I don't expect many new issues from this though.  Furthermore, I
> assume that some of the packages that passed through NEW might not
> build with gcc 4.1 but again that should not be a show stopper.
> I found one serious bug in 4.1.1 though (#370308) which needs to be
> fixed before 4.1 can be the default (since it produces a bogus error
> on some Perl headers which get included by many packages).  Matthias
> is aware of this and is (I think) working on an update already (one
> single patch needs to be reverted).  gcc-defaults can be changed after
> a new 4.1.1 package enters the archive.

4.1.1-2 uploaded. I'll upload a new gcc-defaults with a build
dependency on that version tomorrow.

> In terms of architectures, there wasn't terribly much feedback from
> the porters.

right, we did see most feedback from the non-release architectures.

> However, I have done full archive rebuilds on a number
> of the architectures in the meantime and it looks good.  I might have
> missed a bug or two since I didn't investigate each error in detail
> this time but I think most errors I ignored where generic build
> problems or something related to the new X.
> The status, as far as I'm concerned, is the following:
>    arch            status    tests
> -------------------------------------------------
> +  alpha           good      full archive rebuilt
> +  amd64           good      full archive rebuilt
> ?  arm             unknown   untested
> ?  hppa            would benefit from 4.1 (abi); untested

it requires 4.1 (or removal of 4.1); architecture specific abi
changes.  all packages depending on libgcc2 need to be rebuilt. once
the required packages are rebuilt, libgcc4 will need to conflict with

> +  i386            good      full archive rebuilt
> ?  ia64            unknown   untested
> +  mips            good      full archive rebuilt
> +  mipsel          see mips
> +  powerpc         good      full archive rebuilt
> -  hurd-i386       didn't build until recently; fixed in 4.1.1
> ?  kfreebsd-i386   they'd like to have 4.1
> +  m68k            4.1 fixes many compiler bugs
> ?  s390            unknown   untested
> +  sparc           good      full archive rebuilt
> -- 
> Martin Michlmayr
> http://www.cyrius.com/

Reply to: