[gentoo-science] sage -> numpy -> lapack

[gentoo-science] sage -> numpy -> lapack

[Thomas Kahle]

[-- Attachment #1: Type: text/plain, Size: 285 bytes --]

Hi,

The way that sage depends on lapack and numpy is broken at the moment.
Can somebody look at bug 320669 ?
At least here on x86 it is not possible to get a working sage at the
moment. Let me know if there is anything I can test or do.

Thanks
Thomas

-- 
Thomas Kahle


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 380 bytes --]

Re: [gentoo-science] sage -> numpy -> lapack

[François Bissey]

> Hi,
> 
> The way that sage depends on lapack and numpy is broken at the moment.
> Can somebody look at bug 320669 ?
> At least here on x86 it is not possible to get a working sage at the
> moment. Let me know if there is anything I can test or do.
> 
this is a bit more subtle than that Thomas. I have 2 x86 machines,
one with problems and one without. Christopher doesn't have problems
on his x86 machine either. 
Why is this bug not affecting amd64?

Francois

Re: [gentoo-science] sage -> numpy -> lapack

[Thomas Kahle]

[-- Attachment #1: Type: text/plain, Size: 969 bytes --]

On 06/30/2010 10:51 AM, François Bissey wrote:
>> Hi,
>>
>> The way that sage depends on lapack and numpy is broken at the moment.
>> Can somebody look at bug 320669 ?
>> At least here on x86 it is not possible to get a working sage at the
>> moment. Let me know if there is anything I can test or do.
>>
> this is a bit more subtle than that Thomas. I have 2 x86 machines,
> one with problems and one without. Christopher doesn't have problems
> on his x86 machine either. 
> Why is this bug not affecting amd64?

Now I'm curious about lapack and friends. It seems that atlas is the
current way to go for sage, but I found that in various places
"reference" is eselected on my system.
Can somebody with a working system post a list of which implementations
should be eselected for which package?
Maybe someone should update this here too:
http://www.gentoo.org/proj/en/science/blas-lapack.xml

Thomas


> 
> Francois


-- 
Thomas Kahle


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 380 bytes --]

Re: [gentoo-science] sage -> numpy -> lapack

[François Bissey]

> On 06/30/2010 10:51 AM, François Bissey wrote:
> >> Hi,
> >> 
> >> The way that sage depends on lapack and numpy is broken at the moment.
> >> Can somebody look at bug 320669 ?
> >> At least here on x86 it is not possible to get a working sage at the
> >> moment. Let me know if there is anything I can test or do.
> > 
> > this is a bit more subtle than that Thomas. I have 2 x86 machines,
> > one with problems and one without. Christopher doesn't have problems
> > on his x86 machine either.
> > Why is this bug not affecting amd64?
> 
> Now I'm curious about lapack and friends. It seems that atlas is the
> current way to go for sage, but I found that in various places
> "reference" is eselected on my system.
> Can somebody with a working system post a list of which implementations
> should be eselected for which package?
> Maybe someone should update this here too:
> http://www.gentoo.org/proj/en/science/blas-lapack.xml
> 
After checking there is currently only one place where we have made a
hard requirement on atlas and that is in sage-core where we ask for 
lapack-atlas. I have removed the code eselecting lapack-atlas a few days
ago. If you can test that everything build with reference (you may have to 
edit the ebuild) I may make that dependency into virtual/lapack instead.
Have you built linbox with blas/cblas/lapack reference as well?

Cheers,
Francois