[3.2] Reconsider patches for bison 1.50?

[3.2] Reconsider patches for bison 1.50?

[Zack Weinberg]

This is a quote from the Debian package changelog for gcc 3.2:

  * FTBS: With the switch to bison-1.50 (and 1.75), gcc-3.2 fails to build from
    source on Debian unstable systems. This is fixed in gcc HEAD, but not on
    the current release branch. 
    HELP NEEDED:
    - check what is missing from the patches in debian/patches/bison.dpatch.
      This is a backport of the bison related patches, but showing regressions
      in the gcc testsuite, so it cannot be applied.
    - build gcc using byacc (bootstrap currently fails using byacc).
    - build bison-1.35 in it's own package (the current 1.35-3 package fails
      to build form source).
    - and finally ask upstream to backport the patch to the branch. It's not
      helpful not beeing able to follow the stable branch. Maybe we should
      just switch to gcc HEAD as BSD does ...
    As a terrible workaround, build the sources from CVS first on a machine,
    with bison-1.35 installed, then package the tarball, so the bison
    generated files are not rebuilt.

I think we should reconsider not backporting the patches for bison 1.50
to the 3.2 branch.

zw

Re: [3.2] Reconsider patches for bison 1.50?

[Matthias Klose]

Zack Weinberg writes:
> This is a quote from the Debian package changelog for gcc 3.2:
> 
>   * FTBS: With the switch to bison-1.50 (and 1.75), gcc-3.2 fails to build from
>     source on Debian unstable systems. This is fixed in gcc HEAD, but not on
>     the current release branch. 
>     HELP NEEDED:
>     - check what is missing from the patches in debian/patches/bison.dpatch.
>       This is a backport of the bison related patches, but showing regressions
>       in the gcc testsuite, so it cannot be applied.
>     - build gcc using byacc (bootstrap currently fails using byacc).
>     - build bison-1.35 in it's own package (the current 1.35-3 package fails
>       to build form source).
>     - and finally ask upstream to backport the patch to the branch. It's not
>       helpful not beeing able to follow the stable branch. Maybe we should
>       just switch to gcc HEAD as BSD does ...
>     As a terrible workaround, build the sources from CVS first on a machine,
>     with bison-1.35 installed, then package the tarball, so the bison
>     generated files are not rebuilt.
> 
> I think we should reconsider not backporting the patches for bison 1.50
> to the 3.2 branch.

btw, I noticed the very same regressions on HEAD hppa-linux, but not
on HEAD i386-linux.

Re: [3.2] Reconsider patches for bison 1.50?

[Mark Mitchell]

> I think we should reconsider not backporting the patches for bison 1.50
> to the 3.2 branch.

OK; let's reconsider.  Please point me at the patches.

Thanks,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: [3.2] Reconsider patches for bison 1.50?

[Matthias Klose]

Mark Mitchell writes:
> > I think we should reconsider not backporting the patches for bison 1.50
> > to the 3.2 branch.
> 
> OK; let's reconsider.  Please point me at the patches.

http://gcc.gnu.org/ml/gcc-patches/2002-10/msg00787.html

Re: [3.2] Reconsider patches for bison 1.50?

[Mark Mitchell]

--On Monday, October 21, 2002 10:18:06 AM +0200 Matthias Klose 
<doko@cs.tu-berlin.de> wrote:

> Mark Mitchell writes:
>> > I think we should reconsider not backporting the patches for bison 1.50
>> > to the 3.2 branch.
>>
>> OK; let's reconsider.  Please point me at the patches.
>
> http://gcc.gnu.org/ml/gcc-patches/2002-10/msg00787.html

Very well; go ahead and apply those patches.

Thanks,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com