Sage 3.3.rc0 released

[mabshoff]

Hello folks,

here is rc0, a day or so late. The tarball and its friends have been
relaxing for the last 12 hours in the usual place since I build tested
it once to make 100% there were no new build failures before
announcing and then fell asleep, so if you didn't watch IRC you lost
out to be the first person to report an issue. At least two people did
an upgrade and did not report any test failures, so things are looking
pretty good. :)

Anyway, as usual the bits are in

http://sage.math.washington.edu/home/mabshoff/release-cycles-3.3/

Compared over 3.3.a6 we fixed the MPIR "make check" failure with
specific XCode releases and also merged a bunch of speedup fixes in
the graph as well as matrix code. Some other long standing issues that
were being help up by recent Pynac changes made it into this release
after the issues with Pynac were resolved. There were more fixes for
the notebook and interact, improvements to the number field code, as
well as various random other improvements and bug fixes.

Note that I did merge the gzip patch from #2113 that turned on
compression for the notebook per default since I forgot to actually
merge the patch itself in 3.3.alpha6 after I added the new
twisted.spkg. I do thank Dan Drake for noticing and pointing it out to
me in IRC :)

From here on there will be an rc1, but the 3.3 release needs to get
out soon. The number of open issues and blockers has gone down to the
point where the only true blocker is the FC 9/10 out of memory issue
at startup with linSingular which so far does ironically not even have
a ticket. I will take care of that shortly. But there are still
roughly 13 tickets open against 3.3 and for simplicity below a a copy
& paste from my watch list. Feel free to known a item off the list
today :)

Cheers,

Michael

=============================================================================

mabshoff's watch list for 3.3:

No ticket:

* FC 9/10 out of memory at startup due to libSingular [this one is
the only true blocker]
* libSingular + mallox=system minpoly and Rational() dtor blowup

Unresolved blocker:

#4942: find_root() is broken when interval borders cannot be evaluated
(Mike looked at it, it seems like a scipy issue)
#5100: worksheets: can't empty the trash (safari only?) (no lead)

Needs spkg/patch:

#4666: Michael Abshoff: Make -bdist create canonical binary names
#5216: Michael Abshoff: Update bzip2 to 1.0.5 release (Security)
#5217: Michael Abshoff: update libpng to 1.2.34 (Security)
#5218: Michael Abshoff: Update Pyhton to 2.5.4 (Security)
#5219: Michael Abshoff: Build ATLAS in dist mode with SSE2 only (make
sure to set appropriate sage-flags)

Needs doctest fix:

#4752: Josh Kantor: list_plot3d crashes sage with some exact input
[Reviewed by Mike Hansen, William Stein, Michael Abshoff]

Needs a patch the interfaces/gap.py - this one is important for the
Debian people:

#3337: David Joyner, Michael Abshoff: Upgrade gap-guava to new
4.4.12/3.9 release [Reviewed by Michael Abshoff, Carl Witty]

Needs rereview:

#5098: Yann Laigle-Chapuy, John Cremona: Pollard rho algorithm for
generic discrete logarithm
#5112: Yann Laigle-Chapuy: generic Pollard lambda algorithm (has
dependency on #5098)

Needs review:

#5106: Robert Bradshaw: preparse bug with time and generator
assignment naming
#5104: Robert Bradshaw: setup.py dependency checking issues

=============================================================================

Merged in Sage 3.3.rc0:

#2898: Jason Grout, Nick Alexander, Carl Witty: Coerce integral float
and RDF to Integers [Reviewed by Robert Bradshaw, Mike Hansen]
#3214: Alex Ghitza, Burcin Erocal: uniformise the behaviour of gcd for
rational numbers [Reviewed by John Cremona, Robert Bradshaw]
#4721: Paul Butler: Indefinite integration for piecewise functions
[Reviewed by David Joyner]
#5101: Carl Witty: more types for sage_input: vectors, matrices, etc.
[Reviewed by Jason Grout]
#5142: John Palmieri: speed up elementary_divisors for sparse integer
matrices [Reviewed by Alex Ghitza]
#5185: John Palmieri: is_zero is broken for sparse vectors [Reviewed
by Carl Witty]
#5194: Mike Hansen: add option to turn off automatic updates for an
interact [Reviewed by John Perry]
#5199: Carl Witty: new symbolics can treat floats as integers
inappropriately [Reviewed by Burcin Erocal]
#5205: Michael Abshoff: Set "# -*- coding: utf-8 -*-" encoding for
sage/server/notebook/template.py [Reviewed by Mike Hansen]
#5207: Alex Ghitza: remove unnecessary use of symbolics in doctests in
weierstrass_morphism.py [Reviewed by John Cremona]
#5208: Rob Beezer: Differing behavior for matrix left_kernel vs.
right_kernel [Reviewed by Mike Hansen, Michael Abshoff]
#5209: Michael Abshoff: Sage 3.3.a6: numerical noise in sage/plot/
plot.py on OpenSUSE 11.1 64 bit [Reviewed by Marshall Hampton]
#5210: Bill Hart, Gonzalo Tornaria, Michael Abshoff: gmp-
mpir-0.9.rc3: make check failure on various OSX boxen [Reviewed by
Nick Alexander]
#5215: Michael Abshoff: Remove ipython1-20070130.spkg from Sage
[Reviewed by Mike Hansen]
#5212: Nick Alexander: bug with numbers in names in
sage.structure.parent_gens.normalize_names [Reviewed by David Roe]
#5222: William Stein: wrap FLINTs pseudo-division algorithm for
univariate polynoials over ZZ [Reviewed by Burcin Erocal]
#5224: Robert Miller: speed up isomorphism testing for graphs which
are c_graphs [Reviewed by Michael Abshoff]
#5231: Nick Alexander: make relative number fields lazy [Reviewed by
David Roe]
#5232: William Stein: interact -- major bug in interact ranges due to
str versus repr [Reviewed by Mike Hansen]
#5233: Mike Hansen: improve timings for adjacency_matrix,
weighted_adjacency_matrix, and kirchoff_matrix [Reviewed by Jason
Grout]

[mhampton]

All tests passed on a intel mac pro, running 10.4.11.

-M. Hampton

[mabshoff]

On Feb 11, 12:40 pm, mhampton <hampto...@gmail.com> wrote:
> All tests passed on a intel mac pro, running 10.4.11.

Good.

> -M. Hampton

<SNIP>

> > At least two people did
> > an upgrade and did not report any test failures, so things are looking
> > pretty good. :)

Oh well, it didn't take long, but we do have a load(dump(s)) segfault
in

devel/sage/sage/rings/real_lazy.pyx

See #5234. It happens on 32 as well as 64 bit Linux, one is an
upgrade, the other one a build from scratch, so this is something
serious :(

Cheers,

Michael

[Jaap Spies]

On Fedora 10, 32 bits all tests passed

On Fedora 9, 32 bit the usual mysterious failures, but
passing in a second run.

On Ubuntu 8.10, 32 bits a segfault in devel/sage/sage/rings/real_lazy.pyx
This is now http://trac.sagemath.org/sage_trac/ticket/5234

Jaap

[John H Palmieri]

Hi Michael,

> #5142: John Palmieri: speed up elementary_divisors for sparse integer
> matrices [Reviewed by Alex Ghitza]

It looks like you used the wrong patch for this: please use "5142-
new.patch" from the ticket. (The difference is in the documentation.)

John

[mabshoff]

On Feb 11, 1:59 pm, John H Palmieri <jhpalmier...@gmail.com> wrote:
> Hi Michael,

Hi John,

Yes, you are right. I merged 5142-rebased.patch instead of 5142-
new.patch. So I reverted 5142-rebased.patch and merged 5142-new.patch
in 3.3.rc1. Sorry for the screwup.

>   John

Cheers,

Michael

[David M. Monarres]

On a Macbook with 10.5.6 I get the following doctest failure in
plot.py: (small enough to include in email)

sage -t "devel/sage/sage/plot/plot.py"
**********************************************************************
File "/Users/ayeq/sage-3.3.alpha5/devel/sage/sage/plot/plot.py", line
2582:
sage: generate_plot_points(sin(x), (-pi, pi), randomize=False)
Expected:
[(-3.1415926535897931, -1.2246...e-16), (-2.748893571891069,
-0.38268343236508989), (-2.3561944901923448,
-0.707106781186547...),
(-2.1598449493429825, -0.831469612302545...), (-1.9634954084936207,
-0.92387953251128674), (-1.7671458676442586, -0.98078528040323043),
(-1.5707963267948966, -1.0), (-1.3744467859455345,
-0.98078528040323043), (-1.1780972450961724, -0.92387953251128674),
(-0.98174770424681035, -0.83146961230254524),
(-0.78539816339744828,
-0.707106781186547...), (-0.39269908169872414,
-0.38268343236508978),
(0.0, 0.0), (0.39269908169872414, 0.38268343236508978),
(0.78539816339744828, 0.707106781186547...), (0.98174770424681035,
0.83146961230254524), (1.1780972450961724, 0.92387953251128674),
(1.3744467859455345, 0.98078528040323043), (1.5707963267948966,
1.0),
(1.7671458676442586, 0.98078528040323043), (1.9634954084936207,
0.92387953251128674), (2.1598449493429825, 0.831469612302545...),
(2.3561944901923448, 0.707106781186547...), (2.748893571891069,
0.38268343236508989), (3.1415926535897931, 1.2246...e-16)]
Got:
[(-3.1415926535897931, -1.2246467991473532e-16),
(-2.748893571891069, -0.38268343236508984), (-2.3561944901923448,
-0.70710678118654757), (-2.1598449493429825, -0.83146961230254546),
(-1.9634954084936207, -0.92387953251128674), (-1.7671458676442586,
-0.98078528040323043), (-1.5707963267948966, -1.0),
(-1.3744467859455345, -0.98078528040323043), (-1.1780972450961724,
-0.92387953251128674), (-0.98174770424681035, -0.83146961230254512),
(-0.78539816339744828, -0.70710678118654746), (-0.39269908169872414,
-0.38268343236508978), (0.0, 0.0), (0.39269908169872414,
0.38268343236508978), (0.78539816339744828, 0.70710678118654746),
(0.98174770424681035, 0.83146961230254512), (1.1780972450961724,
0.92387953251128674), (1.3744467859455345, 0.98078528040323043),
(1.5707963267948966, 1.0), (1.7671458676442586, 0.98078528040323043),
(1.9634954084936207, 0.92387953251128674), (2.1598449493429825,
0.83146961230254546), (2.3561944901923448, 0.70710678118654757),
(2.748893571891069, 0.38268343236508984), (3.1415926535897931,
1.2246467991473532e-16)]
**********************************************************************
otherwise everything else passes.

This is from an upgrade from alpha5. (the last version to compile
correctly)

--
David Monarres
dmmon...@gmail.com

"Which is worse: ignorance or apathy? Who knows? Who cares?"

[Dan Drake]

On Wed, 11 Feb 2009 at 12:07PM -0800, mabshoff wrote:
> here is rc0, a day or so late. The tarball and its friends have been
> relaxing for the last 12 hours in the usual place since I build tested
> it once to make 100% there were no new build failures before
> announcing and then fell asleep, so if you didn't watch IRC you lost
> out to be the first person to report an issue. At least two people did
> an upgrade and did not report any test failures, so things are looking
> pretty good. :)

I'm one of the people who did an upgrade, and have now run "make test",
"make ptest", and "make ptestlong" with no persistent failures.

Dan

--
--- Dan Drake <dr...@kaist.edu>
----- KAIST Department of Mathematical Sciences
------- http://mathsci.kaist.ac.kr/~drake

[mabshoff]

On Feb 11, 3:54 pm, Dan Drake <dr...@kaist.edu> wrote:

Hi Dan,

<SNIP>

> I'm one of the people who did an upgrade, and have now run "make test",
> "make ptest", and "make ptestlong" with no persistent failures.

Thanks, but why run all three? "make ptestlong" is all you need since
it does everything and more than the other two. You can always run
ptestlong a couple times if you want to and there is even undocumented
support in ptest to run each doctest n times in a row instead of only
once. What is missing to make that useful is reports that tell you
that $FOO failed x out of y times instead of telling you that it
failed. :)

> Dan

Cheers,

Michael

> --
> ---  Dan Drake <dr...@kaist.edu>
> -----  KAIST Department of Mathematical Sciences
> -------  http://mathsci.kaist.ac.kr/~drake
>

>  signature.asc
> < 1KViewDownload

[mabshoff]

On Feb 11, 3:53 pm, "David M. Monarres" <dmmonar...@gmail.com> wrote:

Hi David,

> On a Macbook with 10.5.6 I get the following doctest failure in  
> plot.py: (small enough to include in email)

<SNIP>

This is now http://trac.sagemath.org/sage_trac/ticket/5239 and I
should have a patch in a short while.

There is a positively reviewed ticket in trac that fixes a wester.py
problem, but since it is subject to numerical noise it will have to
wait port 3.3 :)

Cheers,

Michael

[Dan Drake]

On Wed, 11 Feb 2009 at 04:07PM -0800, mabshoff wrote:
> On Feb 11, 3:54 pm, Dan Drake <dr...@kaist.edu> wrote:
> > I'm one of the people who did an upgrade, and have now run "make
> > test", "make ptest", and "make ptestlong" with no persistent
> > failures.
>
> Thanks, but why run all three? "make ptestlong" is all you need since
> it does everything and more than the other two. You can always run
> ptestlong a couple times if you want to and there is even undocumented
> support in ptest to run each doctest n times in a row instead of only
> once. What is missing to make that useful is reports that tell you
> that $FOO failed x out of y times instead of telling you that it
> failed. :)

I run all three because different things happen with them. I've noticed
that I tend to get fewer Lisp crashes when I run ptest, which seems
strange -- you'd think things would be more reliable when the system
isn't as heavily loaded, but that's not how it works. I also seem to get
random tests timing out during one flavor of test that don't time out
during other flavors.

I know that there isn't a big difference -- and that there *shouldn't*
be -- but the results of running the different tests aren't identical,
and it makes me feel better to run the different flavors of 'make test'.
So I do. :)

I would love to see support for running each test n times and reporting
the number of failures, if any.

Dan

[John Cremona]

I get one test failure for 3.3.rc0:

sage -t "devel/sage/sage/misc/package.py"

but I think it might have been a problem with my home internet
connection while I was out of the house, as it works fine now.
ubuntu 32-bit.

John

[John Cremona]

On 11 Feb, 20:54, Jaap Spies <j.sp...@hccnet.nl> wrote:

>
> On Ubuntu 8.10, 32 bits a segfault in devel/sage/sage/rings/real_lazy.pyx
> This is nowhttp://trac.sagemath.org/sage_trac/ticket/5234

I had the same as that on here:

masgaj@host-56-150%cat /proc/version
Linux version 2.6.16.60-0.33-smp (geeko@buildhost) (gcc version 4.1.2
20070115 (SUSE Linux)) #1 SMP Fri Oct 31 14:24:07 UTC 2008

John


>
> Jaap