On Mon, Aug 20, 2007 at 08:05:48PM +0100, Marcos Marado wrote:
> On Monday 20 August 2007 13:41, Rainer Dorsch wrote:
>> Am Montag, 20. August 2007 13:25 schrieb Marcos Marado:
> >> On Saturday 18 August 2007 08:18, Paul TBBle Hampson wrote:
> > >> I've uploaded 1.18.1.2 packages, however I have no idea if it will
> > >> build on i386 or amd64, as I've got a patch in there which changes the
> > >> way compiler flags are put together compared to upstream (which also
> > >> changed since 1.18.0.6, hence the patch)
> >> Thank you.
> >> Rainer: are you going to keep doing i386 etch packages and test this for
> >> Paul, or do you want me to try doing that?
>> Marcos,
>> I am just back from vacation and have a backlog of work. So I would
>> probably not get started this week.
>> Please go ahead and give the i386 build a try.
> Looking to SConstruct:
> opts.AddOptions(
> EnumOption('BUILD', 'Set build type', 'releasefordownload',
>
> allowed_values=('debug', 'release', 'releasenoopt', 'releasefordownload')),
> EnumOption('ARCH', 'Set architecture', 'i686',
> allowed_values=('i686', 'powerpc', 'x86_64')),
> EnumOption('BTARGET', 'Set build target', 'server',
> allowed_values=('client', 'server', 'all')),
> BoolOption('DISTCC', 'Enabled distcc', True),
> BoolOption('MOZLIB', 'Enabled llmozlib/mozilla support', True),
> BoolOption('FMOD', 'Enabled FMOD audio support', True),
> BoolOption('GSTREAMER', 'Enabled GStreamer support', True),
> BoolOption('COLORGCC', 'Enabled colorgcc', True),
> EnumOption('GRID', 'Client package\'s default grid', 'default',
>
> allowed_values=('default', 'aditi', 'agni', 'dmz', 'durga', 'firstlook', 'ganga', 'shakti', 'siva', 'soma', 'uma', 'vaak')),
> BoolOption('ELFIO', 'Enabled enhanced backtraces with libELFIO symbol
> extraction support', True),
> BoolOption('STANDALONE', 'Build using system packages (implies
> OPENSOURCE)', False),
> BoolOption('OPENSOURCE', 'Build using only non-proprietary dependencies',
> True), # OPENSOURCE: do not edit this line
> BoolOption('VECTORIZE', 'Enable vectorized routines (SIMD instruction sets
> eg. SSE, Altivec) (requires STANDALONE)', True),
> )
> So, STANDALONE is False, but VECTORIZE (at True) needs STANDALONE... What's
> wrong here?
Nothing. The debian/rules file passes in STANDALONE=True.
This however is broken for the non-STANDALONE case. The reason I made
VECTORIZE depend on STANDALONE was because gcc before gcc4 didn't have
the -f option to enable autovectorisation.
That's not so interesting for a Debian patch, we always build
standalone.
--
Paul "TBBle" Hampson, Paul.Hampson@Pobox.com
Shorter .sig for a more eco-friendly paperless office.
Attachment:
pgp8Wd3svT30x.pgp
Description: PGP signature