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

Re: Cross-compiling for powerpcspe



* David Kuehling | 2010-10-29 12:18:58 [+0200]:

>Hi,
Hi David,

> - Is there a mailinglist around the unofficial powerpcspe Debian port?
>   Just use the powerpc list?
    Currently just the powerpc mailing list. We stay here until we are
    not welcome anymore :) If we (Kyle + I) fix powerpcspe related bugs
    we usually Cc each other. If there are more people which want to
    know about it then we could start to Cc the list.

> - I wonder what's required to get a pre-build powerpcspe
>   cross-toolchain into Emdebian?  Current gcc-4.4 from Sid seems to
>   work fine.  Would be happy to supply patches if somebody could state
>   what else needs to be done.
    You should pick up the source from the unrealeased archive [0]. We
    have some extra patches which we don't get into 4.4 tree because we
    can't proof that it is a regression from 4.3. They are in trunk.

> - Installing a sysroot via apt-cross is somewhat difficult for
>   powerpcspe since it requires installing packages from both 'sid' and
>   'unreleased' suites.  Apt-cross can only use one suite at a time, and
>   doing alternating installs from 'sid' and 'unreleased' requires
>   properly selecting package order and causes all kind of dependency
>   problems.  Any known workaround for that?
    I had a similar problem with debootstrap. For debootstrap I created
    one dist with reprepro and used overrides for "essential" packages.

> - It seems to be impossible to find the source-packages corresponding
>   to the binary packages in debian-ports.  E.g. debian-ports contains
>   gcc-4.4_4.4.4-4+powerpcspe4.5.dsc, however the corresponding
>   gcc-4.4_4.4.4.orig.tar.gz is not present on any debian mirror AFAICS.
>   Did I overlook something?
    They are there if you look into the package folder. The uploads
    into unreleased don't update the source dist i.e. there is none.

As Wolfang mentioned we still have atleast one gcc bug which may produce
wrong code under certain circumstances. Thus you should not use this in
any production environment. A workaround is to add -fno-schedule-insns
(as last paramter) another should be to disable double precision
floating point.

>Thanks for any help,
>cheers,
[0]
http://ftp.de.debian.org/debian-ports/pool-powerpcspe/main/g/gcc-4.4/gcc-4.4_4.4.4-4+powerpcspe4.5.diff.gz

>David

Sebastian


Reply to: