[sage-devel] Re: Fwd: mercurial --> plain text --> mercurial

2011-08-11 Thread Keshav Kini 
Google Groups has a new interface (which looks like Google Reader). They 
seem to have lifted that restriction along with the unveiling of the new 
interface. It's been around for a few months. I guess they're distancing 
themselves from their Usenet roots, i.e. no longer keeping their own 
internal distinction between active threads and archived threads to match 
with "extant" and "already purged" messages on Usenet servers (?).

-Keshav


Join us in #sagemath on irc.freenode.net !

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: Another segfault with ATLAS and large matrices

2011-08-11 Thread leif 
On 12 Aug., 02:15, leif  wrote:
> On 11 Aug., 23:21, Volker Braun  wrote:
>
> > I built a Sage binary distribution on redhawk with atlas-3.8.4.spkg and
> > SAGE_FAT_BINARY=yes. It is at:
>
> >http://boxen.math.washington.edu/home/vbraun/Sage/sage-4.7.1.rc2/dist...
>
> > Please let me know if it fixes the segfaults. Or post your backtrace and
> > processor info if not ;-)
>
> Unfortunately it doesn't, see #11674 for details.

Perhaps disabling multithreading in ATLAS when building binary dists
helps.

I wonder if redhawk's libpthread is somehow incompatible; does it
really run 10.04.*1*?

I have (Ubuntu 10.04.3):

-rwxr-xr-x 1 root root 135745 Jan 21  2011 /lib/libpthread-2.11.1.so
lrwxrwxrwx 1 root root 20 Feb  2  2011 /lib/libpthread.so.0 ->
libpthread-2.11.1.so

-rw-r--r-- 1 root root 40298 Jan 21  2011 /usr/include/pthread.h

So it is clearly younger than that originally shipped with 10.04(.1);
alternatively, we would have to include redhawk's, but that's IMHO a
rather bad idea. Better upgrade redhawk...


-leif

P.S.: I'll attach my pthread.h to #11674.

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: Another segfault with ATLAS and large matrices

2011-08-11 Thread leif 
On 11 Aug., 23:21, Volker Braun  wrote:
> I built a Sage binary distribution on redhawk with atlas-3.8.4.spkg and
> SAGE_FAT_BINARY=yes. It is at:
>
> http://boxen.math.washington.edu/home/vbraun/Sage/sage-4.7.1.rc2/dist...
>
> Please let me know if it fixes the segfaults. Or post your backtrace and
> processor info if not ;-)

Unfortunately it doesn't, see #11674 for details.


-leif

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: Another segfault with ATLAS and large matrices

2011-08-11 Thread Volker Braun 
I built a Sage binary distribution on redhawk with atlas-3.8.4.spkg and 
SAGE_FAT_BINARY=yes. It is at:

http://boxen.math.washington.edu/home/vbraun/Sage/sage-4.7.1.rc2/dist/sage-4.7.1.rc2-fat-x86_64-Linux.tar.gz

Please let me know if it fixes the segfaults. Or post your backtrace and 
processor info if not ;-)

Volker

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: [sage-edu] Geogebra 4 coming

2011-08-11 Thread Bruce Cohen 
Sounds exciting.

Take a look at 

-Bruce 
(on the road)

Sent from thing2

On Aug 11, 2011, at 8:49 AM, kcrisman  wrote:

> I was mentioning Geogebra to a student today, and decided to see
> what's up.
> 
> Perhaps this gives additional incentive for people to work on
> http://trac.sagemath.org/sage_trac/ticket/7489.  Does anyone know what
> happened to Bruce Cohen's code where he took the stuff already there
> and built a working prototype of the Geogebra integration in the
> notebook?  I don't see anything on that ticket.
> 
> 

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: Fwd: mercurial --> plain text --> mercurial

2011-08-11 Thread leif 
Sorry, OT:

On 11 Aug., 16:56, Keshav Kini  wrote:
> Sorry for the necropost.

How did you manage that?

At least Google's web interface refused a mailing list reply (to a
thread of Feb/March 2011!) recently; it only gave me the option to
directly reply to the author (without any explanation btw., there was
just the "Reply to author" link at the bottom, and "More options"
didn't show more either.)


-leif

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: size of a Sage install

2011-08-11 Thread leif 
On 11 Aug., 19:26, leif  wrote:
> * I for example get the following for a Sage 4.7.1.rc0 and 4.7.rc2
> installation, respectively:
>
> 950M    devel/sage-11577
> 105M    devel/sage-11602
> 238M    devel/sage-11602-v2
> 288M    devel/sage-11602-v3
> 409M    devel/sage-11602-v4
> 119M    devel/sage-11618
> 102M    devel/sage-11618-v2
> 277M    devel/sage-main
>
> 1,2G            devel/sage-11475v4
> 305M    devel/sage-11475v5
> 85M             devel/sage-main
>
> ... which is IMHO quite strange. Haven't taken a closer look at this
> yet. It's certainly not the Mercurial repository which causes these
> extreme differences; some, but not all might at least partially be due
> to the contents of the doc/ subdirectories.

These are the respective doc/ directories:

$ du -sh devel/sage-*/doc
7,2M  devel/sage-11577/doc
7,2M  devel/sage-11602/doc
7,2M  devel/sage-11602-v2/doc
7,2M  devel/sage-11602-v3/doc
140M devel/sage-11602-v4/doc
7,2M  devel/sage-11618/doc
7,2M  devel/sage-11618-v2/doc
7,9M  devel/sage-main/doc

$ (cd /home/leif/Sage/sage-4.7.rc2/ && du -sh devel/sage-*/doc)
231M devel/sage-11475v4/doc
16M  devel/sage-11475v5/doc
11M  devel/sage-main/doc

So only part of the size variations are due to different documentation
built, and these don't make up the differences in whole.


-leif

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: size of a Sage install

2011-08-11 Thread leif 
On 11 Aug., 15:17, John Cremona  wrote:
> In this case there's no chance of making any such changes, and I do
> not really need to anyway.  All that was important from the point of
> view of Sage was to check that there had not been a 4-fold increase in
> size from 4.6.2 to 4.7;  which there was not.

FWIW, I noticed strange size differences between different Sage
branches ("clones", in subdirectories of devel/).

That's obviously not John's problem, but we [I?] should inspect this
further. If someone else wants to, just do

du -sh devel/sage-*

in case you happen to have (or use) clones at all.*


-leif

_
* I for example get the following for a Sage 4.7.1.rc0 and 4.7.rc2
installation, respectively:

950Mdevel/sage-11577
105Mdevel/sage-11602
238Mdevel/sage-11602-v2
288Mdevel/sage-11602-v3
409Mdevel/sage-11602-v4
119Mdevel/sage-11618
102Mdevel/sage-11618-v2
277Mdevel/sage-main

1,2Gdevel/sage-11475v4
305Mdevel/sage-11475v5
85M devel/sage-main

... which is IMHO quite strange. Haven't taken a closer look at this
yet. It's certainly not the Mercurial repository which causes these
extreme differences; some, but not all might at least partially be due
to the contents of the doc/ subdirectories.

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Geogebra 4 coming

2011-08-11 Thread kcrisman 
I was mentioning Geogebra to a student today, and decided to see
what's up.

In brief, a lot!  So this is an FYI to the Sage community.

Here, Hohenwarter's abstract on a talk for the upcoming Geogebra 4 at
their big conference - including GeogebraTube
http://ggbconference2011.pbworks.com/w/page/40058998/Hohenwarter-Keynote

A good overview by a fan of the new features - some are very nice,
obviously easier in Java than at the command line :)
http://mathandmultimedia.com/2011/06/12/the-geogebra-4-0-sneak-peek-series/

Especially for the sage-edu folks:  In the meantime Geogebra has a
whole series of "Institutes" where you can get training and even
'certification'.  Here is a sample website for such a place:
http://geogebraithaca.wikispaces.com/

I personally think this is something that would be great for Sage, but
our target clientele is (at the moment) significantly smaller, I think
- mostly college and up, and not the huge HS market.

There is also now a Midwest Geogebra Journal, Geogebra is now (at
least partly) on Google Code...  Anyway, a lot going on, worth
pointing out.

Perhaps this gives additional incentive for people to work on
http://trac.sagemath.org/sage_trac/ticket/7489.  Does anyone know what
happened to Bruce Cohen's code where he took the stuff already there
and built a working prototype of the Geogebra integration in the
notebook?  I don't see anything on that ticket.

- kcrisman

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: Fwd: mercurial --> plain text --> mercurial

2011-08-11 Thread Keshav Kini 
Sorry for the necropost. After talking to William I posted a patch at #3052, 
if someone wants to take a look. It doesn't export the repository as diffs, 
but as the more simplistic patches Mercurial actually uses internally 
(consisting of a series, per changeset per file, of ranges to delete and 
blocks of data to replace them with), collated in JSON.

-Keshav


Join us in #sagemath on irc.freenode.net !

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Re: [sage-devel] Re: size of a Sage install

2011-08-11 Thread John Cremona 
In this case there's no chance of making any such changes, and I do
not really need to anyway.  All that was important from the point of
view of Sage was to check that there had not been a 4-fold increase in
size from 4.6.2 to 4.7;  which there was not.

John


>
> In case it is, that's easy to solve without changing the [main]
> filesystem's characteristics, assuming you're using Linux:
>
> [To the owner or administrator of the affected box:]
>
> # Create a sufficiently large file to create a filesystem in (here 4
> GB):
>
> dd if=/dev/zero of=/wherever/sage-filesystem bs=1024 count=4M
>
> # Create a filesystem (here ext4) within that file:
> # (You might get prompted whether to continue because it's not a block
> device.)
>
> mkfs -t ext4 /wherever/sage-filesystem
>
> # Create a directory to later mount that filesystem on (usually in /
> mnt or /media):
>
> mkdir /mnt/sage-filesystem
>
> # Mount the new filesystem:
>
> mount -t ext4 -o loop /wherever/sage-filesystem /mnt/sage-filesystem
>
> # Move (or copy) the Sage installation into that filesystem:
>
> mv $SAGE_ROOT /mnt/sage-filesystem # or use "cp -pr ..."
>
> # Sage now lives in e.g. /mnt/sage-filesystem/sage-4.7
>
>
> You'll need administrator privileges ("sudo ...") for some of the
> commands above, and depending on your setup, also for (re)mounting the
> Sage filesystem later. (You can change the latter by editing /etc/
> fstab and setting the appropriate options for that filesystem, i.e.,
> add "user", and/or perhaps also "auto".)
>
>
> -leif
>
> --
> To post to this group, send an email to sage-devel@googlegroups.com
> To unsubscribe from this group, send an email to 
> sage-devel+unsubscr...@googlegroups.com
> For more options, visit this group at 
> http://groups.google.com/group/sage-devel
> URL: http://www.sagemath.org
>

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: size of a Sage install

2011-08-11 Thread leif 
On 11 Aug., 14:12, Jeroen Demeyer  wrote:
> On 2011-08-11 12:41, John Cremona wrote:
> > After building Sage the amount of filespace used is (as we all know)
> > rather large.  I was just checking a 4.7 install, which is not used
> > for any development pusposes, and is on a machine which I use but do
> > not own, and where I was just told that the filespace was getting
> > short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
> > is in devel and 8.4GB in local.  Moreover, in devel I find:
>
> > masgaj@hpclogin1%cd devel/
> > masgaj@hpclogin1%du -sh *
> > 256K       old
> > 256K       sage
> > 3.1G       sage-main
> > 256K       sagenb
> > 2.4G       sagenb-main
> > masgaj@hpclogin1%ls sagenb-main/
> > build  go        MANIFEST.in  README.txt         sagenb           sass
> >   setup.cfg  spkg-dist
> > dist   Makefile  PKG-INFO     release_notes.txt  sagenb.egg-info
> > sdist  setup.py   SPKG.txt
> > masgaj@hpclogin1%ls sage-main/
> > build   c_lib  export   MANIFEST.in     module_list.pyc  pull
> > sage       setup.py      spkg-dist
> > bundle  doc    install  module_list.py  PKG-INFO         README.txt
> > sage-push  spkg-delauto  spkg-install
>
> > Can anyone say just why sagenb-main is almost as large as sage-main
> > itself?  And is there a simple way (I looked for a command line option
> > but could not see one) of stripping down a Sage build so it is still
> > fully functional, including all the include files and shared libraries
> > under local/ (which I use for building other stuff) but which takes up
> > rather less than 15GB?
>
> I agree with leif, this is probably an issue with the filesystem you are
> using.  This is on an reiserfs file system:
> $ cd $SAGE_ROOT
> $ du -sh      # On-disk size
> 3.0G    .
> $ du -bsh     # Sum of byte sizes of files
> 2.7G    .

In case it is, that's easy to solve without changing the [main]
filesystem's characteristics, assuming you're using Linux:

[To the owner or administrator of the affected box:]

# Create a sufficiently large file to create a filesystem in (here 4
GB):

dd if=/dev/zero of=/wherever/sage-filesystem bs=1024 count=4M

# Create a filesystem (here ext4) within that file:
# (You might get prompted whether to continue because it's not a block
device.)

mkfs -t ext4 /wherever/sage-filesystem

# Create a directory to later mount that filesystem on (usually in /
mnt or /media):

mkdir /mnt/sage-filesystem

# Mount the new filesystem:

mount -t ext4 -o loop /wherever/sage-filesystem /mnt/sage-filesystem

# Move (or copy) the Sage installation into that filesystem:

mv $SAGE_ROOT /mnt/sage-filesystem # or use "cp -pr ..."

# Sage now lives in e.g. /mnt/sage-filesystem/sage-4.7


You'll need administrator privileges ("sudo ...") for some of the
commands above, and depending on your setup, also for (re)mounting the
Sage filesystem later. (You can change the latter by editing /etc/
fstab and setting the appropriate options for that filesystem, i.e.,
add "user", and/or perhaps also "auto".)


-leif

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Re: [sage-devel] size of a Sage install

2011-08-11 Thread John Cremona 
Thanks for all the comments and suggestions.  Two different machines,
and both with IBM gpfs filesystems (both machines are big clusters,
with 1000 and 3300 nodes respectively).

On the older machine I see (in SAGE_ROOT with 4.6.2):

masgaj@fe1%du -sh
3.8G.
masgaj@fe1%du -bsh
2.5G.

while on the newer one (with 4.7):
masgaj@hpclogin1%du -sh
15G .
masgaj@hpclogin1%du -bsh
3.1G.

So if I understand correctly the actual number of bytes in all the
files is not that much larger in the 4.7 build, it's the block size of
the files which has the effect of the files taking up more room on the
second (newer) machine.  I don't know why they are different, but it's
not Sage to blame!

John

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Re: [sage-devel] size of a Sage install

2011-08-11 Thread Jeroen Demeyer 
On 2011-08-11 12:41, John Cremona wrote:
> After building Sage the amount of filespace used is (as we all know)
> rather large.  I was just checking a 4.7 install, which is not used
> for any development pusposes, and is on a machine which I use but do
> not own, and where I was just told that the filespace was getting
> short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
> is in devel and 8.4GB in local.  Moreover, in devel I find:
> 
> masgaj@hpclogin1%cd devel/
> masgaj@hpclogin1%du -sh *
> 256K  old
> 256K  sage
> 3.1G  sage-main
> 256K  sagenb
> 2.4G  sagenb-main
> masgaj@hpclogin1%ls sagenb-main/
> build  goMANIFEST.in  README.txt sagenb   sass
>   setup.cfg  spkg-dist
> dist   Makefile  PKG-INFO release_notes.txt  sagenb.egg-info
> sdist  setup.py   SPKG.txt
> masgaj@hpclogin1%ls sage-main/
> build   c_lib  export   MANIFEST.in module_list.pyc  pull
> sage   setup.py  spkg-dist
> bundle  docinstall  module_list.py  PKG-INFO README.txt
> sage-push  spkg-delauto  spkg-install
> 
> Can anyone say just why sagenb-main is almost as large as sage-main
> itself?  And is there a simple way (I looked for a command line option
> but could not see one) of stripping down a Sage build so it is still
> fully functional, including all the include files and shared libraries
> under local/ (which I use for building other stuff) but which takes up
> rather less than 15GB?

I agree with leif, this is probably an issue with the filesystem you are
using.  This is on an reiserfs file system:
$ cd $SAGE_ROOT
$ du -sh  # On-disk size
3.0G.
$ du -bsh # Sum of byte sizes of files
2.7G.

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Re: [sage-devel] size of a Sage install

2011-08-11 Thread Alex Ghitza 
Oops, of course I forgot to actually add the cc.  Done now.

On Thu, Aug 11, 2011 at 9:34 PM, Alex Ghitza  wrote:
> Dear John,
>
> On Thu, Aug 11, 2011 at 8:41 PM, John Cremona  wrote:
>> After building Sage the amount of filespace used is (as we all know)
>> rather large.  I was just checking a 4.7 install, which is not used
>> for any development pusposes, and is on a machine which I use but do
>> not own, and where I was just told that the filespace was getting
>> short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
>> is in devel and 8.4GB in local.  Moreover, in devel I find:
>
> For the last year or so I have been using the prebuilt version of Sage
> that comes packaged in Arch Linux.  Somehow the maintainer of this
> version (Thomas Dziedzic) manages to create a fully functional
> self-contained Sage distribution that takes about 1.9GB for Sage 4.7.
>  I'm ccing Thomas in case he has some words of wisdom to offer (and I
> take the opportunity to thank him for updating the package so quickly
> after each new Sage release).
>
>
> Best,
> Alex

--
Alex Ghitza -- Lecturer in Mathematics -- The University of Melbourne
-- http://aghitza.org

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Re: [sage-devel] size of a Sage install

2011-08-11 Thread Alex Ghitza 
Dear John,

On Thu, Aug 11, 2011 at 8:41 PM, John Cremona  wrote:
> After building Sage the amount of filespace used is (as we all know)
> rather large.  I was just checking a 4.7 install, which is not used
> for any development pusposes, and is on a machine which I use but do
> not own, and where I was just told that the filespace was getting
> short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
> is in devel and 8.4GB in local.  Moreover, in devel I find:

For the last year or so I have been using the prebuilt version of Sage
that comes packaged in Arch Linux.  Somehow the maintainer of this
version (Thomas Dziedzic) manages to create a fully functional
self-contained Sage distribution that takes about 1.9GB for Sage 4.7.
 I'm ccing Thomas in case he has some words of wisdom to offer (and I
take the opportunity to thank him for updating the package so quickly
after each new Sage release).


Best,
Alex


--
Alex Ghitza -- Lecturer in Mathematics -- The University of Melbourne
-- http://aghitza.org

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: size of a Sage install

2011-08-11 Thread leif 
On 11 Aug., 13:09, leif  wrote:
> On 11 Aug., 12:50, John Cremona  wrote:
>
> > On Thu, Aug 11, 2011 at 11:41 AM, John Cremona  
> > wrote:
> > > After building Sage the amount of filespace used is (as we all know)
> > > rather large.  I was just checking a 4.7 install, which is not used
> > > for any development pusposes, and is on a machine which I use but do
> > > not own, and where I was just told that the filespace was getting
> > > short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
> > > is in devel and 8.4GB in local.
>
> > PS: A sage-4.6.2 install was only 3.8G.  Now I know that 4.7 was alot
> > better than 4.6.2, but FOUR times the size?
>
> On the same machine (and filesystem)?
>
> The size of devel/sage (which is just a symbolic link) indicates that
> the block size of the filesystem where the "huge" Sage installation
> lives is quite large, such that even very small files take up a lot of
> disk space.
>
> There are also differences between different architectures (i.e. 32-
> bit vs. 64-bit, and CISC vs. RISC, e.g. SPARC), but I don't think that
> matters here.
>
> -leif
>
> P.S.:
>
> You can safely delete devel/old/ etc. and $SAGE_ROOT/spkg/build/ in
> case it exists.

P.P.S.:

Do a

tar cf - $SAGE_ROOT | wc -c

and compare that (the number of *bytes* a tarball would take) to what
you get by

du -sh $SAGE_ROOT  # or du -sch $SAGE_ROOT/*


If the former is equally large, there are certainly things that could
be deleted.


-leif

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: size of a Sage install

2011-08-11 Thread leif 


On 11 Aug., 12:50, John Cremona  wrote:
> On Thu, Aug 11, 2011 at 11:41 AM, John Cremona  wrote:
> > After building Sage the amount of filespace used is (as we all know)
> > rather large.  I was just checking a 4.7 install, which is not used
> > for any development pusposes, and is on a machine which I use but do
> > not own, and where I was just told that the filespace was getting
> > short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
> > is in devel and 8.4GB in local.
>
> PS: A sage-4.6.2 install was only 3.8G.  Now I know that 4.7 was alot
> better than 4.6.2, but FOUR times the size?

On the same machine (and filesystem)?

The size of devel/sage (which is just a symbolic link) indicates that
the block size of the filesystem where the "huge" Sage installation
lives is quite large, such that even very small files take up a lot of
disk space.

There are also differences between different architectures (i.e. 32-
bit vs. 64-bit, and CISC vs. RISC, e.g. SPARC), but I don't think that
matters here.


-leif

P.S.:

You can safely delete devel/old/ etc. and $SAGE_ROOT/spkg/build/ in
case it exists.

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: Another segfault with ATLAS and large matrices

2011-08-11 Thread leif 
On 10 Aug., 08:45, leif  wrote:
> On 10 Aug., 07:42, leif  wrote:
>
> > I don't know yet if this is again with a *binary* Sage distribution:

It is.


> >http://trac.sagemath.org/sage_trac/ticket/11674
>
> I can reproduce this error with the *binary* Sage 4.7 distribution on
> Ubuntu 10.04.3 x86_64 (Core2, Penryn).
>
> Note that it was built on redhawk, an AMD Opteron 8439 SE; I btw.
> don't know why the outdated 10.04.1 release is used on that.


Could someone perhaps try building a binary distribution (with
SAGE_FAT_BINARY=yes and) with the new ATLAS spkg from #10226 [1]
(which isn't merged into 4.7.1) on redhawk to see if that fixes the
problem?

I think it's quite odd to just tell people "compile from source" when
they get segfaults with the binary distribution, and in case the new
ATLAS spkg should fix this, we IMHO shouldn't continue to provide
broken binary dists for 4.7.1.

Another try would be to prepare (at least) the Ubuntu 10.04.* 64-bit
binary distribution on a different machine.


-leif

[1] http://trac.sagemath.org/sage_trac/ticket/10226

> > See also
>
> >http://trac.sagemath.org/sage_trac/ticket/11581
>
> > for a previous one, and also
>
> >http://groups.google.com/group/sage-support/browse_thread/thread/4b42...

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: size of a Sage install

2011-08-11 Thread John Cremona 
On Thu, Aug 11, 2011 at 11:41 AM, John Cremona  wrote:
> After building Sage the amount of filespace used is (as we all know)
> rather large.  I was just checking a 4.7 install, which is not used
> for any development pusposes, and is on a machine which I use but do
> not own, and where I was just told that the filespace was getting
> short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
> is in devel and 8.4GB in local.

PS: A sage-4.6.2 install was only 3.8G.  Now I know that 4.7 was alot
better than 4.6.2, but FOUR times the size?

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] size of a Sage install

2011-08-11 Thread John Cremona 
After building Sage the amount of filespace used is (as we all know)
rather large.  I was just checking a 4.7 install, which is not used
for any development pusposes, and is on a machine which I use but do
not own, and where I was just told that the filespace was getting
short, and found that the SAGE_ROOT was taking up 15GB of which 5.5GB
is in devel and 8.4GB in local.  Moreover, in devel I find:

masgaj@hpclogin1%cd devel/
masgaj@hpclogin1%du -sh *
256Kold
256Ksage
3.1Gsage-main
256Ksagenb
2.4Gsagenb-main
masgaj@hpclogin1%ls sagenb-main/
build  goMANIFEST.in  README.txt sagenb   sass
  setup.cfg  spkg-dist
dist   Makefile  PKG-INFO release_notes.txt  sagenb.egg-info
sdist  setup.py   SPKG.txt
masgaj@hpclogin1%ls sage-main/
build   c_lib  export   MANIFEST.in module_list.pyc  pull
sage   setup.py  spkg-dist
bundle  docinstall  module_list.py  PKG-INFO README.txt
sage-push  spkg-delauto  spkg-install

Can anyone say just why sagenb-main is almost as large as sage-main
itself?  And is there a simple way (I looked for a command line option
but could not see one) of stripping down a Sage build so it is still
fully functional, including all the include files and shared libraries
under local/ (which I use for building other stuff) but which takes up
rather less than 15GB?

John

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[sage-devel] Re: website/download stats

2011-08-11 Thread Harald Schilly 
On Wednesday, August 10, 2011 8:15:39 PM UTC+2, leif wrote:
>
> There's not necessarily a direct relation between the download speed 
> and how "near" (geographically) a server is. 
>

Yes, that's also my impression based on some tests and so on. Especially 
when you are on a slower connection, a good server setup is much more 
crucial - because of the network congestion - than physical location. The 
best way to get a fast download without testing anything is using this 
"metalink". I'm still hoping browsers will support it natively some day in 
the future (given, that it is a RFC now).

About the mirror page, yes, I know the layout is not optimal but at least 
you should almost never end up at a broken link. A script pings each mirror 
every 10 minutes and checks if it is alive. I'll try to make some changes 
based on your ideas ...

H

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org