Re: RFC sci-mathematics/axiom

Hi Francois and everyone,

> Have you tried to contact upstream (Tim Daly) to see if being packaged in
> a distro is something they are interested in.
> I personally would last rite it and possibly revisit the situation every so 
> often. If it doesn't even build I am not sure there is a point in keeping it
> in the tree.

I'm in contact with Tim Daly now, who seems very interested in getting
axiom to work on Gentoo. The main problem that I see here, and also with
the developers of Macaulay 2, is that upstream often does not feel any
urge to unsplit their build systems. They just build internal copies of
whatever they need.  So here with axiom it is that the already internal
gcl builds an internal gmp and upstream would like it better to apply
our gentoo patches to that internal gmp instead of making it work with
system wide gmp.  I find it hard to communicate that internal copies are
evil(TM).  How do you guys handle that ?

Cheers,
Thomas

--

-- 
Thomas Kahle

Re: RFC sci-mathematics/axiom

Thomas Kahle wrote:

> I'm in contact with Tim Daly now, who seems very interested in getting
> axiom to work on Gentoo. The main problem that I see here, and also with
> the developers of Macaulay 2, is that upstream often does not feel any
> urge to unsplit their build systems. They just build internal copies of
> whatever they need.  So here with axiom it is that the already internal
> gcl builds an internal gmp and upstream would like it better to apply
> our gentoo patches to that internal gmp instead of making it work with
> system wide gmp.  I find it hard to communicate that internal copies are
> evil(TM).  How do you guys handle that ?

You can always point to one of the numerous flameeyes blog post [1]
against bundled libraries, but I'm sure you can come up with a more
concise explanation.

[1] http://blog.flameeyes.eu/2009/03/23/bundling-libraries-the-curse-of-the-ancients

--

-- 
Sébastien

Re: RFC sci-mathematics/axiom

On 09:09 Fri 01 Oct     , Thomas Kahle wrote:
> Hi Francois and everyone,
> 
> > Have you tried to contact upstream (Tim Daly) to see if being packaged in
> > a distro is something they are interested in.
> > I personally would last rite it and possibly revisit the situation every so 
> > often. If it doesn't even build I am not sure there is a point in keeping it
> > in the tree.
> 
> I'm in contact with Tim Daly now, who seems very interested in getting
> axiom to work on Gentoo. The main problem that I see here, and also with
> the developers of Macaulay 2, is that upstream often does not feel any
> urge to unsplit their build systems. They just build internal copies of
> whatever they need.  So here with axiom it is that the already internal
> gcl builds an internal gmp and upstream would like it better to apply
> our gentoo patches to that internal gmp instead of making it work with
> system wide gmp.  I find it hard to communicate that internal copies are
> evil(TM).  How do you guys handle that ?

You may communicate that they are simply a requirement for many Linux 
distributions, and leave it at that. One way to satisfy both sides might 
be to have the bundled libraries optionally built. Many developers of 
packages with specialized dependencies want to make it easier to install 
their code, so they want to have the option available (and likely on by 
default). What I recommend the package should do is:

- probe the system to autodetect whether the dependency is installed
- allow manual overrides via configure flags
- if not found, then build/install the bundled version

sci-mathematics/maxima fails to build

Re: sci-mathematics/maxima fails to build

> Hi group!
> 
> I've got some problems installing sci-mathematics/maxima in both my Funtoo
> boxes.
> 
> http://paste.pocoo.org/show/271383/
> 
> The issue seems to be similar to :
> http://github.com/cschwan/sage-on-gentoo/issues/issue/14, and the
> work-around to redefining LDFLAGS works as well. However, something is
> wrong with the way maxima deals with the LDFLAGS.
> 
OK I will have a look, it may be something I overlooked in my patches, it used 
to work here, so it may be caused by a slight change somewhere in the system
since then.

If you want maxima for sage you don't have to use the version located in the 
science overlay. It was a proof of concept for some patches. The version in 
portage will work just as well.

Francois

Re: sci-mathematics/maxima fails to build

On   10/05/10 - 18:57:56, François Bissey wrote:
> > Hi group!
> >
> > I've got some problems installing sci-mathematics/maxima in both my  
> Funtoo
> > boxes.
> >
> > http://paste.pocoo.org/show/271383/
> >
> > The issue seems to be similar to :
> > http://github.com/cschwan/sage-on-gentoo/issues/issue/14, and the
> > work-around to redefining LDFLAGS works as well. However, something  
> is
> > wrong with the way maxima deals with the LDFLAGS.
> >
> OK I will have a look, it may be something I overlooked in my  
> patches, it used
> to work here, so it may be caused by a slight change somewhere in the  
> system
> since then.
> 
> If you want maxima for sage you don't have to use the version located  
> in the
> science overlay. It was a proof of concept for some patches. The  
> version in
> portage will work just as well.
> 5.20.1-r2
> Francois
> 
For what it's worth, on my amd64 I'm unable to build the portage  
maxima-5.20.1-r2 either with

LDFLAGS="-Wl,-01 -Wl,--sort-common -Wl,--as-needed"

There is the same

make[1]: *** [binary-ecl/maxima] Error 1

In fact ecls will not even configure with the indicated LDFLAGS.  
Deleting "-Wl,--sort-common" allows ecls as well as maxima to be built.

Steve

Re: sci-mathematics/maxima fails to build

> On   10/05/10 - 18:57:56, François Bissey wrote:
> > > Hi group!
> > > 
> > > I've got some problems installing sci-mathematics/maxima in both my
> > 
> > Funtoo
> > 
> > > boxes.
> > > 
> > > http://paste.pocoo.org/show/271383/
> > > 
> > > The issue seems to be similar to :
> > > http://github.com/cschwan/sage-on-gentoo/issues/issue/14, and the
> > > work-around to redefining LDFLAGS works as well. However, something
> > 
> > is
> > 
> > > wrong with the way maxima deals with the LDFLAGS.
> > 
> > OK I will have a look, it may be something I overlooked in my
> > patches, it used
> > to work here, so it may be caused by a slight change somewhere in the
> > system
> > since then.
> > 
> > If you want maxima for sage you don't have to use the version located
> > in the
> > science overlay. It was a proof of concept for some patches. The
> > version in
> > portage will work just as well.
> > 5.20.1-r2
> > Francois
> 
> For what it's worth, on my amd64 I'm unable to build the portage
> maxima-5.20.1-r2 either with
> 
> LDFLAGS="-Wl,-01 -Wl,--sort-common -Wl,--as-needed"
> 
> There is the same
> 
> make[1]: *** [binary-ecl/maxima] Error 1
> 
> In fact ecls will not even configure with the indicated LDFLAGS.
> Deleting "-Wl,--sort-common" allows ecls as well as maxima to be built.
> 
OK I tested both:
LDFLAGS="-Wl,-01 -Wl,--sort-common -Wl,--as-needed" emerge -1v --nodeps maxima
LDFLAGS="-Wl,-01 -Wl,--sort-common -Wl,--as-needed" emerge -1v --nodeps ecls
which build respectively maxima-5.20.1-r3 and ecls-10.4.1, on my x86 system 
and it merged fine.
Does it work in your 32bits chroot Steve?

Francois

Re: sci-mathematics/maxima fails to build

Hey, what do I know!

On x86 maxima doesn't respect LDFLAGS! And I remember talking with David Kirby 
on sage-devel that actually ecls doesn't either.
But from the logs from JF, on amd64 the LDFLAGS are duplicated and the problem
is that in the failing call one of them is still in double quotes which I 
think is the problem.

Francois

Re: sci-mathematics/maxima fails to build

On   10/05/10 - 22:08:59, François Bissey wrote:
> > On   10/05/10 - 18:57:56, François Bissey wrote:
> > > > Hi group!
> > > >
> > > > I've got some problems installing sci-mathematics/maxima in  
> both my
> > >
> > > Funtoo
> > >
> > > > boxes.
> > > >
> > > > http://paste.pocoo.org/show/271383/
> > > >
> > > > The issue seems to be similar to :
> > > > http://github.com/cschwan/sage-on-gentoo/issues/issue/14, and  
> the
> > > > work-around to redefining LDFLAGS works as well. However,  
> something
> > >
> > > is
> > >
> > > > wrong with the way maxima deals with the LDFLAGS.
> > >
> > > OK I will have a look, it may be something I overlooked in my
> > > patches, it used
> > > to work here, so it may be caused by a slight change somewhere in  
> the
> > > system
> > > since then.
> > >
> > > If you want maxima for sage you don't have to use the version  
> located
> > > in the
> > > science overlay. It was a proof of concept for some patches. The
> > > version in
> > > portage will work just as well.
> > > 5.20.1-r2
> > > Francois
> >
> > For what it's worth, on my amd64 I'm unable to build the portage
> > maxima-5.20.1-r2 either with
> >
> > LDFLAGS="-Wl,-01 -Wl,--sort-common -Wl,--as-needed"
> >
> > There is the same
> >
> > make[1]: *** [binary-ecl/maxima] Error 1
> >
> > In fact ecls will not even configure with the indicated LDFLAGS.
> > Deleting "-Wl,--sort-common" allows ecls as well as maxima to be  
> built.
> >
> OK I tested both:
> LDFLAGS="-Wl,-01 -Wl,--sort-common -Wl,--as-needed" emerge -1v  
> --nodeps maxima
> LDFLAGS="-Wl,-01 -Wl,--sort-common -Wl,--as-needed" emerge -1v  
> --nodeps ecls
> which build respectively maxima-5.20.1-r3 and ecls-10.4.1, on my x86  
> system
> and it merged fine.
> Does it work in your 32bits chroot Steve?
> 
> Francois
> 
François,

Both of the above fail in my 32bit chroot. maxima will not build with  
the same "binary-ecls/maxima" error and ecls will not pass the  
configure step.

Steve

Re: sci-mathematics/maxima fails to build

> François,
> 
> Both of the above fail in my 32bit chroot. maxima will not build with  
> the same "binary-ecls/maxima" error and ecls will not pass the  
> configure step.

Could you share the log for ecls? And the useflags as well.

Francois