Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[E. Madison Bray]

On Fri, Feb 22, 2019 at 12:51 AM Samuel Lelievre
 wrote:
>
>
>
> Wed 2019-02-20 17:49:57 UTC+1, E. Madison Bray:
> >
> > * Some have linked to the spkg-configure keyword on Trac, but there is
> > not to my knowledge an umbrella task ticket for adding support for
> > using more packages from the system (implemented by providing
> > spkg-configure.m4 files for those packages). I think that would be a
> > good idea, especially for prioritizing that work. I'll work on making
> > such a ticket now.
>
> I see you made one now; link for reference:
>
> - Sage Trac ticket 27330
>   Try to use as many system packages as possible
>   https://trac.sagemath.org/ticket/27330
>
> Thanks!

Yes, I had meant to follow up here, but I was sort of in a rush when I
was putting that together.  Thank you for mentioning it here.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Samuel Lelievre]

Wed 2019-02-20 17:49:57 UTC+1, E. Madison Bray:
>
> * Some have linked to the spkg-configure keyword on Trac, but there is 
> not to my knowledge an umbrella task ticket for adding support for 
> using more packages from the system (implemented by providing 
> spkg-configure.m4 files for those packages). I think that would be a 
> good idea, especially for prioritizing that work. I'll work on making 
> such a ticket now. 

I see you made one now; link for reference:

- Sage Trac ticket 27330
  Try to use as many system packages as possible
  https://trac.sagemath.org/ticket/27330

Thanks!

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[E. Madison Bray]

On Tue, Feb 19, 2019 at 6:41 PM Thierry  wrote:
>
> Hi,
>
> On Mon, Feb 18, 2019 at 04:07:07PM +0100, E. Madison Bray wrote:
> [...]
> > Fixing that's something I've been working toward for practically as
> > long as I've been working on the project, and we've been making real
> > progress lately, which you are encouraged to join in and help with...
>
> Perhaps could you provide a task ticket, plus some pointers to readable
> documentation, howtos, etc.

I have several task tickets for several aspects of this, but a lot of
it is a bit disorganized just because there's so much to do.  Some
major ones are:

* For upgrading the installation method for all packages to use
DESTDIR: https://trac.sagemath.org/ticket/24024

* There are still some tickets linked to from there that have been
awaiting review for a long time, including
https://trac.sagemath.org/ticket/25035 and
https://trac.sagemath.org/ticket/25049

* Many packages should still have an spkg-legacy-uninstall:
https://trac.sagemath.org/ticket/25140

* There is a broader task ticket about getting a `make install`
working for Sage.  Some disagree with this approach entirely, and I
agree there are some arguments for/against it, though I think if
nothing else it motivates many changes that are worth making for their
own sake: https://trac.sagemath.org/ticket/21495

* Some have liked to the spkg-configure keyword on Trac, but there is
not to my knowledge an umbrella task ticket for adding support for
using more packages from the system (implemented by providing
spkg-configure.m4 files for those packages).  I think that would be a
good idea, especially for prioritizing that work.  I'll work on making
such a ticket now.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Samuel Lelièvre]

Tue 2019-02-19 18:13 UTC, Dima:
>
> Trac keyword spkg-configure gives a list of relevant tickets.

Direct link:
https://trac.sagemath.org/query?order=id&desc=1&keywords=~spkg-configure

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Dima Pasechnik]

Trac keyword spkg-configure gives a list of relevant tickets.

On Tue, 19 Feb 2019 18:41 Thierry  Hi,
>
> On Mon, Feb 18, 2019 at 04:07:07PM +0100, E. Madison Bray wrote:
> [...]
> > Fixing that's something I've been working toward for practically as
> > long as I've been working on the project, and we've been making real
> > progress lately, which you are encouraged to join in and help with...
>
> Perhaps could you provide a task ticket, plus some pointers to readable
> documentation, howtos, etc.
>
> Ciao,
> Thierry
>
> --
> You received this message because you are subscribed to the Google Groups
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-devel+unsubscr...@googlegroups.com.
> To post to this group, send email to sage-devel@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Thierry]

Hi,

On Mon, Feb 18, 2019 at 04:07:07PM +0100, E. Madison Bray wrote:
[...]
> Fixing that's something I've been working toward for practically as
> long as I've been working on the project, and we've been making real
> progress lately, which you are encouraged to join in and help with...

Perhaps could you provide a task ticket, plus some pointers to readable
documentation, howtos, etc.

Ciao,
Thierry

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[E. Madison Bray]

On Sun, Feb 17, 2019 at 1:50 PM Emmanuel Charpentier
 wrote:
> Is that possible for all of the current packages in Sage-the distribution ? 
> Depending on one set (or possibly a coice of sets) of numerical libraries 
> would allow us to get rid of them in Sage-the-distribution, which should 
> lighten it considerably...

That is the plan, but "all" will be an uphill battle, just in terms of
sheer numbers, though that is why we are focusing primarily on
low-level and/or hard to build and/or common system-level packages
first.

Of course, one thing sage-the-distribution does well is ensuring a
compatible cross-section of versions for all of Sage's dependencies,
which is not easy to do, and in the past has not been done on common
Linux distributions (but significant progress has been made there).
The downside to that of course is that Sage ends up only being built
and tested on a very narrow, fine-tuned set of dependency versions.
This makes it easier to build/run consistently, but also makes it more
fragile to run on more arbitrary configurations (and I would argue
more fragile in general).

For that reason I think sage-the-distribution will continue to serve a
purpose (even if not in the same form it's in currently, such as if we
switch over to just using conda).  But for a *lot* of the
dependencies, especially on reasonably modern OS distributions, it is
quite possible to get those dependencies from the system and we just
don't.  Fixing that's something I've been working toward for
practically as long as I've been working on the project, and we've
been making real progress lately, which you are encouraged to join in
and help with...

>>> On Sat, 16 Feb 2019 13:40 Emmanuel Charpentier >> wrote:
>
> Trac#27303 needs review.
>
> Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :
>>
>>
>>
>> > On 16/02/2019, at 22:40, Emmanuel Charpentier  
>> > wrote:
>> >
>> > I see (modulo the legendary readability of m4, which has been compared 
>> > to line noise or a cat's nap on the keyboard...).
>> >
>>
>> Any language is a bit like that when you don’t know it to be fair.
>> And you can produce line noise in most programming languages
>> but I’d pick perl and lisp over m4 for that.
>>
>> Do open a ticket for sqlite upgrade, there may be other interesting
>> reasons to upgrade.
>>
>> There may already be a ticket for making sqlite optional and if
>> you think m4 is looks like line noise you probably should leave
>> it to the pros.
>
> --
> You received this message because you are subscribed to the Google Groups 
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to sage-devel+...@googlegroups.com.
> To post to this group, send email to sage-...@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.
>>>
>>> --
>>> You received this message because you are subscribed to the Google Groups 
>>> "sage-devel" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an 
>>> email to sage-devel+...@googlegroups.com.
>>> To post to this group, send email to sage-...@googlegroups.com.
>>> Visit this group at https://groups.google.com/group/sage-devel.
>>> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups 
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to sage-devel+unsubscr...@googlegroups.com.
> To post to this group, send email to sage-devel@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

A couple of remarks :

Le dimanche 17 février 2019 13:17:12 UTC+1, Dima Pasechnik a écrit :
>
>
>
> On Sun, 17 Feb 2019 12:48 Emmanuel Charpentier   wrote:
>
>>
>>
>> Le dimanche 17 février 2019 09:34:10 UTC+1, Dima Pasechnik a écrit :
>>>
>>> Our plan is to enable the use of the system R install in Sage - for this 
>>> we need to understand whether there would be any problems for rpy2 if R 
>>> used a different, from the one in Sage, Blas library.
>>>
>>
We should be cautious with this one :

0.8.2 is our current reference because, AFAICT, it is the last version of 
rpy2 that supports Python2. If/when our plans to switch to Python3 succed 
(at last !), we should explore its newer versions. And switch to them if 
practical. But their requirements for numerics may vary...

Quite probably : In order to use the rjags 
>>  package 
>> (interface to the JAGS  Bayesian 
>> modelling sampler, I had to use the Sage shell in order to compile and 
>> install a version of JAGS in $SAGE_LOCAL : the system JAGS choked on 
>> BLAS/LAPACK calls not available in Sage's libraries.
>>
>> Things would be much simpler if Sage used the systemwide numeric 
>> libraries. I *think* that this is possible when these libraries are ATLAS : 
>> pointing SAGE_ATLAS_LIB to the systemwide ATLAS library makes Sage use 
>> them. ISTR that Erik Madison Bray recommended this to compile Sage on 
>> Cygwin (but I lost the reference to the relevant Sage Wiki post...). 
>>
>> My question was to know if this is possible when openblas provides the 
>> systemwide numerical libraries...
>>
>
> Not at the moment, but there are plans to make it possible.
> One should do more spkg-configure.m4 along the lines of #27212, where this 
> is done for gmp and mpir.
>

Is that possible for all of the current packages in Sage-the distribution ? 
Depending on one set (or possibly a coice of sets) of numerical libraries 
would allow us to get rid of them in Sage-the-distribution, which should 
lighten it considerably...

On Sat, 16 Feb 2019 13:40 Emmanuel Charpentier > wrote:
>>
>>> Trac#27303  needs review.

 Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :
>
>
>
> > On 16/02/2019, at 22:40, Emmanuel Charpentier <
> emanuel.c...@gmail.com> wrote: 
> > 
> > I see (modulo the legendary readability of m4, which has been 
> compared to line noise or a cat's nap on the keyboard...). 
> > 
>
> Any language is a bit like that when you don’t know it to be fair. 
> And you can produce line noise in most programming languages 
> but I’d pick perl and lisp over m4 for that. 
>
> Do open a ticket for sqlite upgrade, there may be other interesting 
> reasons to upgrade. 
>
> There may already be a ticket for making sqlite optional and if 
> you think m4 is looks like line noise you probably should leave 
> it to the pros.

 -- 
 You received this message because you are subscribed to the Google 
 Groups "sage-devel" group.
 To unsubscribe from this group and stop receiving emails from it, send 
 an email to sage-devel+...@googlegroups.com.
 To post to this group, send email to sage-...@googlegroups.com.
 Visit this group at https://groups.google.com/group/sage-devel.
 For more options, visit https://groups.google.com/d/optout.

>>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "sage-devel" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to sage-devel+...@googlegroups.com .
>> To post to this group, send email to sage-...@googlegroups.com 
>> .
>> Visit this group at https://groups.google.com/group/sage-devel.
>> For more options, visit https://groups.google.com/d/optout.
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Indeed.

Thanks for coming to the rescue of my aging neurons...

Le dimanche 17 février 2019 13:24:12 UTC+1, Samuel Lelievre a écrit :
>
> Sun 2019-02-17 12:48:52 UTC+1, Emmanuel Charpentier:
> >
> > Things would be much simpler if Sage used the systemwide numeric
> > libraries. I *think* that this is possible when these libraries are 
> ATLAS:
> > pointing SAGE_ATLAS_LIB to the systemwide ATLAS library makes
> > Sage use them. ISTR that Erik Madison Bray recommended this to
> > compile Sage on Cygwin (but I lost the reference to the relevant
> > Sage Wiki post...). 
>
> You might be referring to this page on the Sage Trac Wiki:
>
> https://trac.sagemath.org/wiki/Cygwin64Port#BuildingSage
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Samuel Lelievre]

Sun 2019-02-17 12:48:52 UTC+1, Emmanuel Charpentier:
>
> Things would be much simpler if Sage used the systemwide numeric
> libraries. I *think* that this is possible when these libraries are ATLAS:
> pointing SAGE_ATLAS_LIB to the systemwide ATLAS library makes
> Sage use them. ISTR that Erik Madison Bray recommended this to
> compile Sage on Cygwin (but I lost the reference to the relevant
> Sage Wiki post...). 

You might be referring to this page on the Sage Trac Wiki:

https://trac.sagemath.org/wiki/Cygwin64Port#BuildingSage

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Timo Kaufmann]

Am Sonntag, 17. Februar 2019 12:48:52 UTC+1 schrieb Emmanuel Charpentier:
>
>
>
> Le dimanche 17 février 2019 09:34:10 UTC+1, Dima Pasechnik a écrit :
>>
>> Our plan is to enable the use of the system R install in Sage - for this 
>> we need to understand whether there would be any problems for rpy2 if R 
>> used a different, from the one in Sage, Blas library.
>>
>
> Quite probably : In order to use the rjags 
>  package 
> (interface to the JAGS  Bayesian 
> modelling sampler, I had to use the Sage shell in order to compile and 
> install a version of JAGS in $SAGE_LOCAL : the system JAGS choked on 
> BLAS/LAPACK calls not available in Sage's libraries.
>
> Things would be much simpler if Sage used the systemwide numeric 
> libraries. I *think* that this is possible when these libraries are ATLAS : 
> pointing SAGE_ATLAS_LIB to the systemwide ATLAS library makes Sage use 
> them. ISTR that Erik Madison Bray recommended this to compile Sage on 
> Cygwin (but I lost the reference to the relevant Sage Wiki post...). 
>
> My question was to know if this is possible when openblas provides the 
> systemwide numerical libraries...
>

Setting `LD_DEBUG=libs` shows you which libraries are used at runtime. 
`LD_PRELOAD` can be useful to force the use of a particular blas 
implementation. 
 

>  
>
>> On Sat, 16 Feb 2019 13:40 Emmanuel Charpentier > wrote:
>>
>>> Trac#27303  needs review.
>>>
>>> Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :



 > On 16/02/2019, at 22:40, Emmanuel Charpentier  
 wrote: 
 > 
 > I see (modulo the legendary readability of m4, which has been 
 compared to line noise or a cat's nap on the keyboard...). 
 > 

 Any language is a bit like that when you don’t know it to be fair. 
 And you can produce line noise in most programming languages 
 but I’d pick perl and lisp over m4 for that. 

 Do open a ticket for sqlite upgrade, there may be other interesting 
 reasons to upgrade. 

 There may already be a ticket for making sqlite optional and if 
 you think m4 is looks like line noise you probably should leave 
 it to the pros.
>>>
>>> -- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "sage-devel" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to sage-devel+...@googlegroups.com.
>>> To post to this group, send email to sage-...@googlegroups.com.
>>> Visit this group at https://groups.google.com/group/sage-devel.
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Dima Pasechnik]

On Sun, 17 Feb 2019 12:48 Emmanuel Charpentier <
emanuel.charpent...@gmail.com wrote:

>
>
> Le dimanche 17 février 2019 09:34:10 UTC+1, Dima Pasechnik a écrit :
>>
>> Our plan is to enable the use of the system R install in Sage - for this
>> we need to understand whether there would be any problems for rpy2 if R
>> used a different, from the one in Sage, Blas library.
>>
>
> Quite probably : In order to use the rjags
>  package
> (interface to the JAGS  Bayesian
> modelling sampler, I had to use the Sage shell in order to compile and
> install a version of JAGS in $SAGE_LOCAL : the system JAGS choked on
> BLAS/LAPACK calls not available in Sage's libraries.
>
> Things would be much simpler if Sage used the systemwide numeric
> libraries. I *think* that this is possible when these libraries are ATLAS :
> pointing SAGE_ATLAS_LIB to the systemwide ATLAS library makes Sage use
> them. ISTR that Erik Madison Bray recommended this to compile Sage on
> Cygwin (but I lost the reference to the relevant Sage Wiki post...).
>
> My question was to know if this is possible when openblas provides the
> systemwide numerical libraries...
>

Not at the moment, but there are plans to make it possible.
One should do more spkg-configure.m4 along the lines of #27212, where this
is done for gmp and mpir.


>
>> On Sat, 16 Feb 2019 13:40 Emmanuel Charpentier > wrote:
>>
>>> Trac#27303  needs review.
>>>
>>> Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :



 > On 16/02/2019, at 22:40, Emmanuel Charpentier 
 wrote:
 >
 > I see (modulo the legendary readability of m4, which has been
 compared to line noise or a cat's nap on the keyboard...).
 >

 Any language is a bit like that when you don’t know it to be fair.
 And you can produce line noise in most programming languages
 but I’d pick perl and lisp over m4 for that.

 Do open a ticket for sqlite upgrade, there may be other interesting
 reasons to upgrade.

 There may already be a ticket for making sqlite optional and if
 you think m4 is looks like line noise you probably should leave
 it to the pros.
>>>
>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "sage-devel" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to sage-devel+...@googlegroups.com.
>>> To post to this group, send email to sage-...@googlegroups.com.
>>> Visit this group at https://groups.google.com/group/sage-devel.
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>> --
> You received this message because you are subscribed to the Google Groups
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-devel+unsubscr...@googlegroups.com.
> To post to this group, send email to sage-devel@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Le dimanche 17 février 2019 09:34:10 UTC+1, Dima Pasechnik a écrit :
>
> Our plan is to enable the use of the system R install in Sage - for this 
> we need to understand whether there would be any problems for rpy2 if R 
> used a different, from the one in Sage, Blas library.
>

Quite probably : In order to use the rjags 
 package 
(interface to the JAGS  Bayesian 
modelling sampler, I had to use the Sage shell in order to compile and 
install a version of JAGS in $SAGE_LOCAL : the system JAGS choked on 
BLAS/LAPACK calls not available in Sage's libraries.

Things would be much simpler if Sage used the systemwide numeric libraries. 
I *think* that this is possible when these libraries are ATLAS : pointing 
SAGE_ATLAS_LIB to the systemwide ATLAS library makes Sage use them. ISTR 
that Erik Madison Bray recommended this to compile Sage on Cygwin (but I 
lost the reference to the relevant Sage Wiki post...). 

My question was to know if this is possible when openblas provides the 
systemwide numerical libraries...
 

> On Sat, 16 Feb 2019 13:40 Emmanuel Charpentier   wrote:
>
>> Trac#27303  needs review.
>>
>> Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :
>>>
>>>
>>>
>>> > On 16/02/2019, at 22:40, Emmanuel Charpentier  
>>> wrote: 
>>> > 
>>> > I see (modulo the legendary readability of m4, which has been compared 
>>> to line noise or a cat's nap on the keyboard...). 
>>> > 
>>>
>>> Any language is a bit like that when you don’t know it to be fair. 
>>> And you can produce line noise in most programming languages 
>>> but I’d pick perl and lisp over m4 for that. 
>>>
>>> Do open a ticket for sqlite upgrade, there may be other interesting 
>>> reasons to upgrade. 
>>>
>>> There may already be a ticket for making sqlite optional and if 
>>> you think m4 is looks like line noise you probably should leave 
>>> it to the pros.
>>
>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "sage-devel" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to sage-devel+...@googlegroups.com .
>> To post to this group, send email to sage-...@googlegroups.com 
>> .
>> Visit this group at https://groups.google.com/group/sage-devel.
>> For more options, visit https://groups.google.com/d/optout.
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Dima Pasechnik]

Our plan is to enable the use of the system R install in Sage - for this we
need to understand whether there would be any problems for rpy2 if R used a
different, from the one in Sage, Blas library.

On Sat, 16 Feb 2019 13:40 Emmanuel Charpentier <
emanuel.charpent...@gmail.com wrote:

> Trac#27303  needs review.
>
> Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :
>>
>>
>>
>> > On 16/02/2019, at 22:40, Emmanuel Charpentier 
>> wrote:
>> >
>> > I see (modulo the legendary readability of m4, which has been compared
>> to line noise or a cat's nap on the keyboard...).
>> >
>>
>> Any language is a bit like that when you don’t know it to be fair.
>> And you can produce line noise in most programming languages
>> but I’d pick perl and lisp over m4 for that.
>>
>> Do open a ticket for sqlite upgrade, there may be other interesting
>> reasons to upgrade.
>>
>> There may already be a ticket for making sqlite optional and if
>> you think m4 is looks like line noise you probably should leave
>> it to the pros.
>
> --
> You received this message because you are subscribed to the Google Groups
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-devel+unsubscr...@googlegroups.com.
> To post to this group, send email to sage-devel@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Trac#27303  needs review.

Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :
>
>
>
> > On 16/02/2019, at 22:40, Emmanuel Charpentier  > wrote: 
> > 
> > I see (modulo the legendary readability of m4, which has been compared 
> to line noise or a cat's nap on the keyboard...). 
> > 
>
> Any language is a bit like that when you don’t know it to be fair. 
> And you can produce line noise in most programming languages 
> but I’d pick perl and lisp over m4 for that. 
>
> Do open a ticket for sqlite upgrade, there may be other interesting 
> reasons to upgrade. 
>
> There may already be a ticket for making sqlite optional and if 
> you think m4 is looks like line noise you probably should leave 
> it to the pros.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Le samedi 16 février 2019 10:47:05 UTC+1, François Bissey a écrit :
>
>
>
> > On 16/02/2019, at 22:40, Emmanuel Charpentier  > wrote: 
> > 
> > I see (modulo the legendary readability of m4, which has been compared 
> to line noise or a cat's nap on the keyboard...). 
> > 
>
> Any language is a bit like that when you don’t know it to be fair. 
> And you can produce line noise in most programming languages 
> but I’d pick perl and lisp over m4 for that. 
>

Regexps as used from R aren't bad, either... ;-).
 

> Do open a ticket for sqlite upgrade, there may be other interesting 
> reasons to upgrade. 
>

Done. That builds without a hitch. make ptestlong underway (it will be a 
while : make started by recompiling both pythons. I think that this will 
recompile almost anything...)
 

> There may already be a ticket for making sqlite optional and if 
> you think m4 is looks like line noise you probably should leave 
> it to the pros.


Or *really* learn m4... which is a bit orthogonal to my initial 
preoccupations (evidence-based medicine and dentistry).
 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[François Bissey]

> On 16/02/2019, at 22:40, Emmanuel Charpentier  
> wrote:
> 
> I see (modulo the legendary readability of m4, which has been compared to 
> line noise or a cat's nap on the keyboard...).
> 

Any language is a bit like that when you don’t know it to be fair.
And you can produce line noise in most programming languages
but I’d pick perl and lisp over m4 for that.

Do open a ticket for sqlite upgrade, there may be other interesting
reasons to upgrade.

There may already be a ticket for making sqlite optional and if 
you think m4 is looks like line noise you probably should leave
it to the pros.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Dear François

Le samedi 16 février 2019 10:30:14 UTC+1, François Bissey a écrit :
>
>
>
> > On 16/02/2019, at 22:25, Emmanuel Charpentier  > wrote: 
> > 
> > Pushing to use system sqlite 
> > 
> > What do you mean ? Do you mean removing "our" sqlite and dependig on 
> system's sqlite ? Nice idea... except for Cygwin,and/or WSL... 
> >   
>
> Erik has been pushing that for a number of package. In a configurable way. 
> For example 
>
> https://github.com/sagemath/sage/blob/develop/build/pkgs/bzip2/spkg-configure.m4
>  
>

I see (modulo the legendary readability of m4, which has been compared to 
line noise or a cat's nap on the keyboard...).

And, ndeed, it should also apply to a lot of "our" libraries, and possibly 
"our" interpreters (ideally, we should be able to reduce 
Sage-the-distribution to Sage-the-library...).

But this is a secondary goal. Some systems may need "our" sqlite (e. g. 
when the system's sqlite is too old and can't be upgraded for some 
reason...).

So I'll propose a patch to upgrade *our* sqlite, and, possibly, another one 
to render "our" version optional.

Comments ? 

sqlite would make as much sense to take from the system as something like 
> bzip2. 
>
> François

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[François Bissey]

> On 16/02/2019, at 22:25, Emmanuel Charpentier  
> wrote:
> 
> Pushing to use system sqlite
> 
> What do you mean ? Do you mean removing "our" sqlite and dependig on system's 
> sqlite ? Nice idea... except for Cygwin,and/or WSL...
>  

Erik has been pushing that for a number of package. In a configurable way.
For example 
https://github.com/sagemath/sage/blob/develop/build/pkgs/bzip2/spkg-configure.m4

sqlite would make as much sense to take from the system as something like
bzip2.

François

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Le samedi 16 février 2019 10:20:51 UTC+1, François Bissey a écrit :
>
> Good diagnostics there. I’d recommend 
> https://sqlite.org/2019/sqlite-autoconf-3270100.tar.gz 
> as the sources to use.


Thanks ! I'll try that
 

> Pushing to use system sqlite


What do you mean ? Do you mean removing "our" sqlite and dependig on 
system's sqlite ? Nice idea... except for Cygwin,and/or WSL...
 

> is 
> also an healthy option. 
>
> François 
>
> > On 16/02/2019, at 22:15, Emmanuel Charpentier  > wrote: 
> > 
> > Ahem... 
> > 
> > It turns put that I can't reproduce my *origonal* problem : sustem's 
> upgrade seems to ghave fixed rthe problem. 
> > 
> > Sorry for the noise... 
> > 
> > However, "spdep" still doesn't install, but the problem si totally 
> different : its dependency "sf" doesn't compile due to a missing library 
> call in ... libgdal.so, of all things, which claims : 
> > 
> > /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_origin_name' 
> > /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_table_name' 
> > collect2: error: ld returned 1 exit status 
> > /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_origin_name' 
> > /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_table_name' 
> > collect2: error: ld returned 1 exit status 
> > 
> > This is more explainable: "our" sqlite3 is at 3.22, whereas system's 
> sqlite3 is at 3.26+fossilbc891ac6b (whatever that means...). 
> > 
> > I might try to update "our" sqlite3 *if* I can figure what is the 
> "right" source Debian unstable has 3.27 ; I might try that. 
> > 
> > Hints ? Suggestions ? 
> > 
> > 
> > Le samedi 16 février 2019 10:00:45 UTC+1, François Bissey a écrit : 
> > Can you send us the error message you get when installing sp/spdep? 
> > What version of openblas do you have on your system? The latest beta 
> > of sage uses openblas 0.3.5 which is latest available release. 
> > 
> > François 
> > 
> > > On 16/02/2019, at 21:46, Emmanuel Charpentier  
> wrote: 
> > > 
> > > Dear list, 
> > > 
> > > I started to get strange errors when installing some R packages ("sp" 
> and "spdep", for the record). Ultimately, it boils down to the fact that 
> those libraries need to use library system calls that exist in the system's 
> openblas (these packages do install fine in system's R) but not in "our" 
> openblas. 
> > > 
> > > A possible workaround would be to install Atlas systemwide and to 
> point SAGE_ATLAS_LIB to that. But  using ATLAS has some drawbacks (e. g. 
> necessity to manually recompile in order to optimize BLAS and LAPACK). 
> Hence my question : 
> > > 
> > > Is there a way to install (i. e. compile) Sage using system's BLAS and 
> LAPACK provided by system's openblas ? 
> > > 
> > > PS : another workaround would be ti upgrade "our" openblas to a 
> version supporting the needed calls. I do not know how to do this, and it 
> sems soewhat intricate. 
> > > 
> > > 
> > > -- 
> > > You received this message because you are subscribed to the Google 
> Groups "sage-devel" group. 
> > > To unsubscribe from this group and stop receiving emails from it, send 
> an email to sage-devel+...@googlegroups.com. 
> > > To post to this group, send email to sage-...@googlegroups.com. 
> > > Visit this group at https://groups.google.com/group/sage-devel. 
> > > For more options, visit https://groups.google.com/d/optout. 
> > 
> > 
> > -- 
> > You received this message because you are subscribed to the Google 
> Groups "sage-devel" group. 
> > To unsubscribe from this group and stop receiving emails from it, send 
> an email to sage-devel+...@googlegroups.com . 
> > To post to this group, send email to sage-...@googlegroups.com 
> . 
> > Visit this group at https://groups.google.com/group/sage-devel. 
> > For more options, visit https://groups.google.com/d/optout. 
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[François Bissey]

Good diagnostics there. I’d recommend 
https://sqlite.org/2019/sqlite-autoconf-3270100.tar.gz
as the sources to use. Pushing to use system sqlite is 
also an healthy option.

François

> On 16/02/2019, at 22:15, Emmanuel Charpentier  
> wrote:
> 
> Ahem...
> 
> It turns put that I can't reproduce my *origonal* problem : sustem's upgrade 
> seems to ghave fixed rthe problem.
> 
> Sorry for the noise...
> 
> However, "spdep" still doesn't install, but the problem si totally different 
> : its dependency "sf" doesn't compile due to a missing library call in ... 
> libgdal.so, of all things, which claims :
> 
> /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_origin_name'
> /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_table_name'
> collect2: error: ld returned 1 exit status
> /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_origin_name'
> /usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
> `sqlite3_column_table_name'
> collect2: error: ld returned 1 exit status
> 
> This is more explainable: "our" sqlite3 is at 3.22, whereas system's sqlite3 
> is at 3.26+fossilbc891ac6b (whatever that means...).
> 
> I might try to update "our" sqlite3 *if* I can figure what is the "right" 
> source Debian unstable has 3.27 ; I might try that.
> 
> Hints ? Suggestions ?
> 
> 
> Le samedi 16 février 2019 10:00:45 UTC+1, François Bissey a écrit :
> Can you send us the error message you get when installing sp/spdep? 
> What version of openblas do you have on your system? The latest beta 
> of sage uses openblas 0.3.5 which is latest available release. 
> 
> François 
> 
> > On 16/02/2019, at 21:46, Emmanuel Charpentier  
> > wrote: 
> > 
> > Dear list, 
> > 
> > I started to get strange errors when installing some R packages ("sp" and 
> > "spdep", for the record). Ultimately, it boils down to the fact that those 
> > libraries need to use library system calls that exist in the system's 
> > openblas (these packages do install fine in system's R) but not in "our" 
> > openblas. 
> > 
> > A possible workaround would be to install Atlas systemwide and to point 
> > SAGE_ATLAS_LIB to that. But  using ATLAS has some drawbacks (e. g. 
> > necessity to manually recompile in order to optimize BLAS and LAPACK). 
> > Hence my question : 
> > 
> > Is there a way to install (i. e. compile) Sage using system's BLAS and 
> > LAPACK provided by system's openblas ? 
> > 
> > PS : another workaround would be ti upgrade "our" openblas to a version 
> > supporting the needed calls. I do not know how to do this, and it sems 
> > soewhat intricate. 
> > 
> > 
> > -- 
> > You received this message because you are subscribed to the Google Groups 
> > "sage-devel" group. 
> > To unsubscribe from this group and stop receiving emails from it, send an 
> > email to sage-devel+...@googlegroups.com. 
> > To post to this group, send email to sage-...@googlegroups.com. 
> > Visit this group at https://groups.google.com/group/sage-devel. 
> > For more options, visit https://groups.google.com/d/optout. 
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups 
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to sage-devel+unsubscr...@googlegroups.com.
> To post to this group, send email to sage-devel@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Ahem...

It turns put that I can't reproduce my *origonal* problem : sustem's 
upgrade seems to ghave fixed rthe problem.

Sorry for the noise...

However, "spdep" still doesn't install, but the problem si totally 
different : its dependency "sf" doesn't compile due to a missing library 
call in ... libgdal.so, of all things, which claims :

/usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
`sqlite3_column_origin_name'
/usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
`sqlite3_column_table_name'
collect2: error: ld returned 1 exit status
/usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
`sqlite3_column_origin_name'
/usr/bin/ld: /usr/lib/libgdal.so: undefined reference to 
`sqlite3_column_table_name'
collect2: error: ld returned 1 exit status

This is more explainable: "our" sqlite3 is at 3.22, whereas system's 
sqlite3 is at 3.26+fossilbc891ac6b (whatever that means...).

I might try to update "our" sqlite3 *if* I can figure what is the "right" 
source Debian unstable has 3.27 ; I might try that.

Hints ? Suggestions ?


Le samedi 16 février 2019 10:00:45 UTC+1, François Bissey a écrit :
>
> Can you send us the error message you get when installing sp/spdep? 
> What version of openblas do you have on your system? The latest beta 
> of sage uses openblas 0.3.5 which is latest available release. 
>
> François 
>
> > On 16/02/2019, at 21:46, Emmanuel Charpentier  > wrote: 
> > 
> > Dear list, 
> > 
> > I started to get strange errors when installing some R packages ("sp" 
> and "spdep", for the record). Ultimately, it boils down to the fact that 
> those libraries need to use library system calls that exist in the system's 
> openblas (these packages do install fine in system's R) but not in "our" 
> openblas. 
> > 
> > A possible workaround would be to install Atlas systemwide and to point 
> SAGE_ATLAS_LIB to that. But  using ATLAS has some drawbacks (e. g. 
> necessity to manually recompile in order to optimize BLAS and LAPACK). 
> Hence my question : 
> > 
> > Is there a way to install (i. e. compile) Sage using system's BLAS and 
> LAPACK provided by system's openblas ? 
> > 
> > PS : another workaround would be ti upgrade "our" openblas to a version 
> supporting the needed calls. I do not know how to do this, and it sems 
> soewhat intricate. 
> > 
> > 
> > -- 
> > You received this message because you are subscribed to the Google 
> Groups "sage-devel" group. 
> > To unsubscribe from this group and stop receiving emails from it, send 
> an email to sage-devel+...@googlegroups.com . 
> > To post to this group, send email to sage-...@googlegroups.com 
> . 
> > Visit this group at https://groups.google.com/group/sage-devel. 
> > For more options, visit https://groups.google.com/d/optout. 
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

Re: [sage-devel] Is there a way to force Sage to use system's openblas ?

[François Bissey]

Can you send us the error message you get when installing sp/spdep?
What version of openblas do you have on your system? The latest beta 
of sage uses openblas 0.3.5 which is latest available release.

François

> On 16/02/2019, at 21:46, Emmanuel Charpentier  
> wrote:
> 
> Dear list,
> 
> I started to get strange errors when installing some R packages ("sp" and 
> "spdep", for the record). Ultimately, it boils down to the fact that those 
> libraries need to use library system calls that exist in the system's 
> openblas (these packages do install fine in system's R) but not in "our" 
> openblas.
> 
> A possible workaround would be to install Atlas systemwide and to point 
> SAGE_ATLAS_LIB to that. But  using ATLAS has some drawbacks (e. g. necessity 
> to manually recompile in order to optimize BLAS and LAPACK). Hence my 
> question :
> 
> Is there a way to install (i. e. compile) Sage using system's BLAS and LAPACK 
> provided by system's openblas ?
> 
> PS : another workaround would be ti upgrade "our" openblas to a version 
> supporting the needed calls. I do not know how to do this, and it sems 
> soewhat intricate.
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups 
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to sage-devel+unsubscr...@googlegroups.com.
> To post to this group, send email to sage-devel@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

[sage-devel] Is there a way to force Sage to use system's openblas ?

[Emmanuel Charpentier]

Dear list,

I started to get strange errors when installing some R packages ("sp" and 
"spdep", for the record). Ultimately, it boils down to the fact that those 
libraries need to use library system calls that exist in the system's 
openblas (these packages do install fine in system's R) but not in "our" 
openblas.

A possible workaround would be to install Atlas systemwide and to point 
SAGE_ATLAS_LIB to that. But  using ATLAS has some drawbacks (e. g. 
necessity to manually recompile in order to optimize BLAS and LAPACK). 
Hence my question :

Is there a way to install (i. e. compile) Sage using system's BLAS and 
LAPACK provided by system's openblas ?

PS : another workaround would be ti upgrade "our" openblas to a version 
supporting the needed calls. I do not know how to do this, and it sems 
soewhat intricate.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.