update: building on skynet machine fulvia (32-bit) almost works

[John H Palmieri]

A 32-bit build on fulvia (Solaris on x86) almost works. To build Sage
4.5.2.rc1: set environment variables as described in <http://
wiki.sagemath.org/skynet>. Use one of the ATLAS spkg's from #9508:
the one that does NOT build liblapack.so. Use the Singular spkg from
#8059, and also patch the Sage library using the patch from that
ticket.

Then type "make ptestlong".

When I do this, I end up with 6 files with doctest failures:

devel/sage/sage/modular/hecke/submodule.py # 1 doctests failed
devel/sage/sage/modular/abvar/abvar.py # 1 doctests failed
devel/sage/sage/lfunctions/sympow.py # 13 doctests failed
devel/sage/sage/stats/hmm/chmm.pyx # 3 doctests failed
devel/sage/sage/schemes/elliptic_curves/ell_rational_field.py # 17
doctests failed
devel/sage/sage/symbolic/expression.pyx # 2 doctests failed

I've posted the log here:

<http://sage.math.washington.edu/home/palmieri/misc/fulvia-
ptestlong.log>.

These failures are of three sorts:

(1) numerical noise, for example:

sage -t -long devel/sage/sage/stats/hmm/chmm.pyx
**********************************************************************
File "/home/palmieri/fulvia/32bit/sage-4.5.2.rc1/devel/sage-main/sage/
stats/hmm/chmm.pyx", line 570:
sage: m.viterbi([0,1,10,10,1])
Expected:
([0, 0, 1, 1, 0], -9.0604285688230899)
Got:
([0, 0, 1, 1, 0], -9.0604285688230917)

and several similar failures from the same file.

(2) small maxima issue with numerical noise and/or how output is
printed:

sage -t -long devel/sage/sage/symbolic/expression.pyx
**********************************************************************
File "/home/palmieri/fulvia/32bit/sage-4.5.2.rc1/devel/sage-main/sage/
symbolic/expression.pyx", li\
ne 4983:
sage: maxima('sinh(1.0)')
Expected:
1.175201193643801
Got:
1.175201193643802
**********************************************************************
File "/home/palmieri/fulvia/32bit/sage-4.5.2.rc1/devel/sage-main/sage/
symbolic/expression.pyx", li\
ne 5088:
sage: maxima('asinh(1.0)')
Expected:
0.881373587019543
Got:
.8813735870195429
**********************************************************************

(3) sympow seems to be really broken, leading to failures in

devel/sage/sage/modular/hecke/submodule.py # 1 doctests failed
devel/sage/sage/modular/abvar/abvar.py # 1 doctests failed
devel/sage/sage/lfunctions/sympow.py # 13 doctests failed
devel/sage/sage/schemes/elliptic_curves/ell_rational_field.py # 17
doctests failed

I get lots of "RuntimeError: failed to compute modular degree" in
these files.

I don't know what the issue is with sympow, but otherwise things are
looking good for 32-bit Solaris on x86.

(64-bit builds have more issues.)

--
John

[Dr. David Kirkby]

On 08/ 5/10 05:13 PM, John H Palmieri wrote:
> A 32-bit build on fulvia (Solaris on x86) almost works. To build Sage
> 4.5.2.rc1: set environment variables as described in<http://
> wiki.sagemath.org/skynet>. Use one of the ATLAS spkg's from #9508:
> the one that does NOT build liblapack.so. Use the Singular spkg from
> #8059, and also patch the Sage library using the patch from that
> ticket.
>
> Then type "make ptestlong".
>
> When I do this, I end up with 6 files with doctest failures:
>
> devel/sage/sage/modular/hecke/submodule.py # 1 doctests failed
> devel/sage/sage/modular/abvar/abvar.py # 1 doctests failed
> devel/sage/sage/lfunctions/sympow.py # 13 doctests failed
> devel/sage/sage/stats/hmm/chmm.pyx # 3 doctests failed
> devel/sage/sage/schemes/elliptic_curves/ell_rational_field.py # 17
> doctests failed
> devel/sage/sage/symbolic/expression.pyx # 2 doctests failed

That's great news John. I've already created a ticket for the last of these
doctest failures

#9689

but the others will need to be done too.

> (3) sympow seems to be really broken, leading to failures in
>
> devel/sage/sage/modular/hecke/submodule.py # 1 doctests failed
> devel/sage/sage/modular/abvar/abvar.py # 1 doctests failed
> devel/sage/sage/lfunctions/sympow.py # 13 doctests failed
> devel/sage/sage/schemes/elliptic_curves/ell_rational_field.py # 17
> doctests failed

> I get lots of "RuntimeError: failed to compute modular degree" in
> these files.
>
> I don't know what the issue is with sympow, but otherwise things are
> looking good for 32-bit Solaris on x86.

sympow is a standalone package - it has no dependencies really. It even avoids
using system maths libraries because of copyright concerns!

I tried building sympow stand-alone (outside of Sage) on my laptop (OpenSolaris)
and it just seem to hang. At least it hang for long enough that I was unwilling
to flatten my battery whilst I was on a train. There's some examples in the file
src/README, but I could not actually get any result in the time I was willing to
let it peg my CPU.

> (64-bit builds have more issues.)

Yes, true. But I think they too can be overcome.

> --
> John
>

Dave