On 7/19/07, Georges Khaznadar <[EMAIL PROTECTED]> wrote:
> William Stein a écrit :
> > (1) Create a monolothic .deb, which installs everything SAGE currently
> > distributes in /opt/sage/, along with a run script /usr/bin/sage.
>
> This stage is OK : the debian package was built finely, but it is done
> using a brutal technique: everything is distributed, even .o files and
> many other useless stuff. The resulting package is bloated (255
> megaBytes!)
Why is it so big? It should be possible to make that package but have
it be about 150MB.
> I don't wish to uplod it anywhere, and prefer going on with stage 2 as
> soon as possible.
I would greatly appreciate it if we could finish stage 1 first, and really
have it something that works well, can be distributed, etc. This would
really be beneficial to many people, and as you have already demonstrated
shouldn't be too difficult.
> S I would like to ask you some questions in order to
> save time and unnecessary trials:
>
> There must be some glue between Sage's core and the other packages like
> Gap et Maxima, etc. As I aim to replace every part of Sage's current
> package which was already packaged independently, by some modification of
> the relevant glue component, so I would like to have some advice about it
> if you can write it.
>
> For example, when I packaged Wims, which is now a package of Debian
> Stable, I had to check that the file
> /var/lib/wims/public_html/bin/maxima which is the wrapper used by Wims
> to call Maxima was compatible with the current debian package already
> existing. Fortunately, there were very few modifications to do inside
> these glue components. Hopefully it may be similar for Sage?
That's unlikely. Usually just using *anything* but the current
version of maxima causes
tons of problems, since they frequently make substantial changes to precision,
special functions, etc. It's a surprisingly actively developed
package. Mathematics
software that wasn't meant to be used in library mode has a very very
complicated
interface that is often not very stable between versions of the
software. That SAGE
packages everything else in a monolithic fashion is perhaps the main technical
reason SAGE is possible at all.
That said, since you're unstoppable :-), each SAGE package (.spkg) is a tar.bz2
ball that has a file spkg-install in it, and a directory "patches"
with all patches
we make against the standard source tarball (actually, patches usually contains
whole files instead of patches -- use diff to get an actual diff if
you want one).
> > (2) Only after having completely mastering (1), move on to considering
> > breaking up SAGE into smaller packages and installing into the standard
> > Debian environment.
>
> So I consider it now. If you want I can release the files
> sage_2.6.orig.tar.gz (105557191 Bytes) and
> sage_2.6-1.diff.gz ( 7633 Bytes)
>
> As you can see, the modification to debianise unproprely sage is very
> lightweight. However I should make a more complete diff file to do it
> following "the debian way" and make these sources yeld a more reasonably
> sized binary package. Hence the call for your advice.
Unfortunately (for your project) at the moment the SAGE distribution itself is
undergoing some fairly extreme modifications as a result of several of the
coding sprint projects at SAGE Days 4. We've added numerous new packages,
and Didier Deshome did a bunch of great work designing a better structure
for SAGE spkg's. The next version of SAGE (2.7.1), which I'll release tomorrow,
will have all spkg's repackaged using this new structure. Also, it will have,
hopefully, build much more reliably than SAGE-2.7, due to switching to
g95 fortran.
I would be extremely grateful if you could download sage-2.7.1 when I release it
tomorrow, or look at
http://sage.math.washington.edu/home/was/sage2.7.1/
which will be pretty close to what sage-2.7.1 will be. If nothing else, I would
like any information about making a Debian SAGE package that you've
come up with so far and would be willing to share. Thanks!
By the way, would you like an account on sage.math.washington.edu? It's
basically a big 16-core 64GB RAM server on which most sage developers
have accounts. It helps greatly with sharing work. If so, write to
me off-list.
-- William
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/
-~----------~----~----~----~------~----~------~--~---
