SAGE 2.10.alpha0 released

[mabshoff]

Hello folks,

and a new merge cycle has been opened. Every time I finish one they
pull me back in again ;)

Anyways, tarball is in

http://sage.math.washington.edu/home/mabshoff/release-cycles-2.10/

Inside that directory you see another directory alpha0 which contains
all updated spkgs as well as all patches. So that way you can keep an
eye on development and just grab patches and spkgs as they get merged.
Once I start alpha1 expect that directory to appear there, too.

As usual I will ask people to do reviews and fix open bugs. Come on
over to the tracker and do your best. alpha0 was mostly about mem leak
fixes, I did discover a whole bunch of new ones when I did an audit
with the help of valgrind of 2.9.3 and with the help of various people
(see below) we fixed a whole bunch of them. We are getting to the
point where the number of known leaks is going toward zero, the main
culprit at the moment is Givaro.

As usual let us know about any issue you encounter. I am catching some
sleep now - see you in the morning.

Cheers,

Michael

Merged in alpha0:

#1092: Willem Jan Palenstijn, Michael Abshoff: small memleaks exposed
by ntl_ZZ_pE
#1093: Michael Abshoff, Willem Jan Palenstijn: small memleaks exposed
by ntl_ZZ_pX.py
#1541: Burcin Erocal: improve PolyBoRi integration
#1544: David Joyner, Rich Morin: SAGE Tutorial nits #2
#1553: Kiran Kedlaya, Michael Abshoff: SCons related build failure
of PolyBoRi on 64-bit RHEL5
#1598: Martin Albrecht: fix SIGSEGV in libSINGULAR interface on
Solaris
#1694: Michael Abshoff, Bill Hart: Update FLINT to 1.05 release
#1701: Martin Albrecht: attempt to clean up currRing if deallocated
#1702: Martin Albrecht: fix memleak in fplll.pyx
#1703: Michael Abshoff: memleak in Singular: one mpz is leaked in
longrat.cc

[Kiran Kedlaya]

Works for me. That means #1553 is solved, though the larger SCons
issue remains in case we bring in any more software that uses it.

On my box (again, that's 64-bit RHEL5), sage -testall passes all
tests.

Kiran

On Jan 7, 9:30 pm, mabshoff <Michael.Absh...@fsmath.mathematik.uni-

[mabshoff]

On Jan 8, 5:04 pm, Kiran Kedlaya <ksk...@gmail.com> wrote:

Hi Kiran,

> Works for me.

Good to know, but that was the intention all the way.

> That means #1553 is solved, though the larger SCons
> issue remains in case we bring in any more software that uses it.

I meant to write the PoLyBoRi team an email with the patch so they can
integrate it upstream. I see no reason not to import PATH from ENV.

> On my box (again, that's 64-bit RHEL5), sage -testall passes all
> tests.

Excellent.

> Kiran

Cheers,

Michael

[Burcin Erocal]

On Tue, 8 Jan 2008 09:34:06 -0800 (PST)
mabshoff <Michael...@fsmath.mathematik.uni-dortmund.de> wrote:

> On Jan 8, 5:04 pm, Kiran Kedlaya <ksk...@gmail.com> wrote:

> > That means #1553 is solved, though the larger SCons
> > issue remains in case we bring in any more software that uses it.
>
> I meant to write the PoLyBoRi team an email with the patch so they can
> integrate it upstream. I see no reason not to import PATH from ENV.

They already know about it, and they will incorporate the fix.

Maybe we should consider patching the SCons package, so that it uses
better default values. I thought this was going to be the solution for
#1656.

Burcin

[mabshoff]

On Jan 8, 8:54 pm, Burcin Erocal <bur...@erocal.org> wrote:
> On Tue, 8 Jan 2008 09:34:06 -0800 (PST)
>

> mabshoff <Michael.Absh...@fsmath.mathematik.uni-dortmund.de> wrote:
> > On Jan 8, 5:04 pm, Kiran Kedlaya <ksk...@gmail.com> wrote:
> > > That means #1553 is solved, though the larger SCons
> > > issue remains in case we bring in any more software that uses it.
>
> > I meant to write the PoLyBoRi team an email with the patch so they can
> > integrate it upstream. I see no reason not to import PATH from ENV.

Hi Burcin,

> They already know about it, and they will incorporate the fix.

Ok, I send the patch to Michael B. today off list, but I hadn't heard
back yet. I also think that that fix is a potential solution for the
Itanium issue that William reported since at least on the box I am
currently playing with a gcc 3.2 gets picked and that doesn't end too
well as you can imagine. I plan to build a new gcc 4.2.2 on that box
tonight and also build 2.10.alpha0 on there to see if the problem is
resolved.

> Maybe we should consider patching the SCons package, so that it uses
> better default values. I thought this was going to be the solution for
> #1656.

Well, #1656 is resolved since we took care of both spkgs that use
SCons. I don't think there are any better default values, especially
since choice A that works on one system will surely not work on
another. We just need to make sure to check any new code that uses
SCons also imports PATH from ENV.

> Burcin

Cheers,

Michael

[Alexander Dreyer]

Hi everybody,

> Ok, I send the patch to Michael B. today off list, but I hadn't heard
> back yet.

I already incorporated the patch from the trac. In fact, I've
rewritten it a little bit, since HOME and LD_LIBRARY_PATH could be
treated likewise, see:
http://polybori.cvs.sourceforge.net/polybori/PolyBoRi/SConstruct?r1=1.8&r2=1.9

> I also think that that fix is a potential solution for the
> Itanium issue that William reported since at least on the box I am
> currently playing with a gcc 3.2 gets picked and that doesn't end too
> well as you can imagine.

Yeah, standalone-PolyBoRi works on Itanium, but for compiling SAGE,
one needs a newer compiler (I have the same problem on the Itanium box
here, but didn't succeed in compiling SAGE yet).

Best regards,
Alexander
PS: Feel free to post these issues and patches directely to the
PolyBoRi-Tracker: http://sourceforge.net/tracker/?group_id=210499

[mabshoff]

On Jan 9, 4:35 pm, Alexander Dreyer

<jan.alexander.dre...@googlemail.com> wrote:
> Hi everybody,> Ok, I send the patch to Michael B. today off list, but I hadn't heard
> > back yet.
>
> I already incorporated the patch from the trac. In fact, I've
> rewritten it a little bit, since HOME and LD_LIBRARY_PATH could be

> treated likewise, see:http://polybori.cvs.sourceforge.net/polybori/PolyBoRi/SConstruct?r1=1...

>
> > I also think that that fix is a potential solution for the
> > Itanium issue that William reported since at least on the box I am
> > currently playing with a gcc 3.2 gets picked and that doesn't end too
> > well as you can imagine.
>
> Yeah, standalone-PolyBoRi works on Itanium, but for compiling SAGE,
> one needs a newer compiler (I have the same problem on the Itanium box
> here, but didn't succeed in compiling SAGE yet).

Ok, I am having similar problems and while I solved most of the
compilation issues I also have some linker issue at startup to solve.
William just send me an email that he never got Sage to work on that
box, so it might be a while before I get this to work.

> Best regards,
> Alexander
> PS: Feel free to post these issues and patches directely to the
> PolyBoRi-Tracker:http://sourceforge.net/tracker/?group_id=210499

Ok, I bookmarked it and will post issues I see there, too.

Cheers,

Michael

[Kate]

Michael,

Sage-2.10.alpha0 builds from source (using gcc-4.2.2)
fine for me on

x86-Linux (pentium4-fc6)
x86_64-Linux (opteron-fc6)

but fails on ia64-Linux (RH Linux) in flint-1.05 with

Testing _fmpz_poly_mul_karatsuba()... ok
Testing _fmpz_poly_mul_karatsuba_trunc()... ok
Testing _fmpz_poly_mul_karatsuba_trunc_left()... ok
Testing _fmpz_poly_mul_KS()... GNU MP: Cannot reallocate memory
(old_size=8 new_size=4294959128)
./spkg-check: line 51: 11240 Aborted (core dumped) ./
fmpz_poly-test

Kate

On Jan 7, 9:30 pm, mabshoff <Michael.Absh...@fsmath.mathematik.uni-
dortmund.de> wrote:

[mabshoff]

On Jan 9, 5:15 pm, Kate <kate01...@gmail.com> wrote:
> Michael,

Hi Kate,

> Sage-2.10.alpha0 builds from source (using gcc-4.2.2)
> fine for me on
>
> x86-Linux (pentium4-fc6)
> x86_64-Linux (opteron-fc6)
>
> but fails on ia64-Linux (RH Linux) in flint-1.05 with
>
> Testing _fmpz_poly_mul_karatsuba()... ok
> Testing _fmpz_poly_mul_karatsuba_trunc()... ok
> Testing _fmpz_poly_mul_karatsuba_trunc_left()... ok
> Testing _fmpz_poly_mul_KS()... GNU MP: Cannot reallocate memory
> (old_size=8 new_size=4294959128)
> ./spkg-check: line 51: 11240 Aborted (core dumped) ./
> fmpz_poly-test

I found the same bug on an Itanium also with gcc 4.2.2. See

https://groups.google.com/group/sage-devel/t/a05734c584f012b

I did run make check on Sage's gmp compiled with that compiler and it
passes with flying colors. Bill Hart is aware of the problem, but
currently is still contemplating what would cause this. It looks like
a potential gmp bug, but so far no one has figured out what is wrong.
He currently needs access to an actual Itanium to track this down, but
it looks like William will give him access shortly.

Another question: I assume we can close http://trac.sagemath.org/sage_trac/ticket/1277
since those two issues you reported there have been resolved. I meant
to ask for confirmation, but I got side tracked over the last couple
days :(

> Kate

Cheers,

Michael

[Kate]

Michael,

> I found the same bug on an Itanium also with gcc 4.2.2. See
>
> https://groups.google.com/group/sage-devel/t/a05734c584f012b

Sorry, missed that.

> Another question: I assume we can closehttp://trac.sagemath.org/sage_trac/ticket/1277

> since those two issues you reported there have been resolved. I meant
> to ask for confirmation, but I got side tracked over the last couple
> days :(

Yes, you may close it. flint built ... just did not pass its tests.

Kate

On Jan 9, 11:23 am, mabshoff <Michael.Absh...@fsmath.mathematik.uni-

dortmund.de> wrote:
> On Jan 9, 5:15 pm, Kate <kate01...@gmail.com> wrote:
>
> > Michael,
>
> Hi Kate,
>
> > Sage-2.10.alpha0 builds from source (using gcc-4.2.2)
> > fine for me on
>
> > x86-Linux (pentium4-fc6)
> > x86_64-Linux (opteron-fc6)
>
> > but fails on ia64-Linux (RH Linux) in flint-1.05 with
>
> > Testing _fmpz_poly_mul_karatsuba()... ok
> > Testing _fmpz_poly_mul_karatsuba_trunc()... ok
> > Testing _fmpz_poly_mul_karatsuba_trunc_left()... ok
> > Testing _fmpz_poly_mul_KS()... GNU MP: Cannot reallocate memory
> > (old_size=8 new_size=4294959128)
> > ./spkg-check: line 51: 11240 Aborted (core dumped) ./
> > fmpz_poly-test
>
> I found the same bug on an Itanium also with gcc 4.2.2. See
>
> https://groups.google.com/group/sage-devel/t/a05734c584f012b
>
> I did run make check on Sage's gmp compiled with that compiler and it
> passes with flying colors. Bill Hart is aware of the problem, but
> currently is still contemplating what would cause this. It looks like
> a potential gmp bug, but so far no one has figured out what is wrong.
> He currently needs access to an actual Itanium to track this down, but
> it looks like William will give him access shortly.
>

> Another question: I assume we can closehttp://trac.sagemath.org/sage_trac/ticket/1277