Re: working around a gcc-4.4 bug
On Thu, Jun 9, 2011 at 5:03 PM, Gerfried Fuchs <rhonda@deb.at> wrote:
> Hey,
>
> * Adam Borowski <kilobyte@angband.pl> [2011-06-09 15:47:11 CEST]:
>> There's a bug in gcc-4.4 that causes an ICE on powerpc on a number of
>> packages, including "crawl" which I just received a report about FTBFSing on
>> backports. This particular ICE can be avoided by building with no
>> optimization.
>>
>> Architectures other than powerpc, and gcc-4.5 or 4.6 work well.
>>
>> What should I do? Does this require a sourceful upload that checks the
>> compiler's version and forces noopt on powerpc? Or is there a better way?
>
> I think in this case an entry in debian/rules that does check with
> dpkg-architecture and turn the switches for this case is the best bet.
>
>> A sourceful upload is not out of question -- I can cut an upstream point
>> release with a bunch of bugfixes so they can get in before Ubuntu's import
>> freeze on June 16. I just wonder what's the least intrusive way.
>
> As you said yourself it only affects gcc 4.4, the only place that needs
> the sourceful upload. Neither Debian testing/unstable nor Ubuntu oneiric
> do have gcc-4.4 as default. It though would also affect Ubuntu
> backports, that might be the only reason to get that into oneiric
> indeed, to ease backports for them too.
You could also notify the maintainers of gcc, so they have some idea
of how to regression test for this particular bug with new versions.
Reply to: