On Sep 23, 4:05 pm, "Georg S. Weber" <[EMAIL PROTECTED]>
wrote:
> Hi,
>
> On 24 Sep., 00:34, mabshoff <[EMAIL PROTECTED]> wrote:
>
> > On Sep 23, 3:27 pm, "Georg S. Weber" <[EMAIL PROTECTED]>
> > wrote:
>
> > > On 23 Sep., 10:07, "John Cremona" <[EMAIL PROTECTED]> wrote:
>
> > <SNIP>
>
> > > I'll open more tickets for other failures, as soon I can say for sure
> > > "it' only timing" and have a new "good" timing constant.
>
> > Please do not open any more tickets, but post the logs somewhere so I
> > can take a long. The timeout for long is 30 minutes, so if the test
> > does not finish it is most likely that the job either got pushed
> > deeply into swap or something else bad happened. "testlong" is not
> > meant for ancient PPC boxen :)
>
> Well, that's a question of how much robustness you want.
> If it does work even on five-year old PPCs, it is less likely to break
> randomly in other "unforseen" circumstances. E.g. porting should be
> the easier, the more robust and stable Sage is ;-)
>
> > > Cheers,
> > > gsw
>
> > Cheers,
>
> > Michael
>
> Here are the log snippets, first three "good ones" to give you a
> timing impression of this old box:
>
> sage -t -long devel/sage/sage/algebras/algebra.py
> [45.0 s]
> sage -t -long devel/sage/sage/algebras/algebra_element.py
> [13.7 s]
> sage -t -long devel/sage/sage/algebras/algebra_ideal.py
> [14.5 s]
>
> <SNIP>
>
> sage -t -long devel/sage/sage/calculus/calculus.py
> **********************************************************************
> File "/Volumes/SageVolume/sage/sage-3.1.3.alpha0/tmp/calculus.py",
> line 2262:
> sage: (z + exp(x)).laplace(x, s)
> Expected:
> z/s + 1/(s - 1)
> Got:
> Maxima crashed -- automatically restarting.
> _tmp_
> **********************************************************************
> 1 items had failures:
> 1 of 26 in __main__.example_64
> ***Test Failed*** 1 failures.
> For whitespace errors, see the file /Volumes/SageVolume/sage/
> sage-3.1.3.alpha0/tmp/.doctest_calculus.py
> [791.4 s]
>
> <SNIP>
>
> sage -t -long devel/sage/sage/crypto/mq/sr.py
> halt 14
>
> error: no more memory
> System 430592k:434784k Appl 425481k/5110k Malloc 4058k/37k Valloc
> 426496k/5073k Pages 106518/106 Regions 833:841
>
> A mysterious error (perphaps a memory error?) occurred, which may have
> crashed doctest.
> [1318.0 s]
>
> <SNIP>
>
> sage -t -long devel/sage/sage/dsage/dist_functions/
> dist_factor.py*************************************************************
> *********
<SNIP>
> For whitespace errors, see the file /Volumes/SageVolume/sage/
> sage-3.1.3.alpha0/tmp/.doctest_dist_factor.py
> [99.3 s]
>
> <SNIP>
Ok, this is a known issue. There are four DSage tickets in trac that
will hopefully be reviewed and merge soon that will fix that issue.
> sage -t -long devel/sage/sage/functions/piecewise.py
<SNIP>
> [1013.0 s]
Maxima goes off to the deep end - not much we can do, but upgrade
clisp :)
> <SNIP>
>
> sage -t -long devel/sage/sage/groups/perm_gps/partn_ref/
> refinement_matrices.pyx*** *** Error: TIMED OUT! *** ***
> *** *** Error: TIMED OUT! *** ***
> [45062.1 s]
Hmm, this is insane, but I think you end up eating large amounts of
swap. Once you compute in swap it is game over. The code does not leak
by the way since we did valgrind it very intensely.
> <SNIP>
>
> sage -t -long devel/sage/sage/interfaces/psage.py
> **********************************************************************
> File "/Volumes/SageVolume/sage/sage-3.1.3.alpha0/tmp/psage.py", line
> 23:
> sage: w = [x('factor(2^%s-1)'% randint(250,310)) for x in v]
> Exception raised:
> Traceback (most recent call last):
> File "/Volumes/SageVolume/sage/sage-3.1.3.alpha0/local/lib/
> python2.5/doctest.py", line 1228, in __run
> compileflags, 1) in test.globs
> File "<doctest __main__.example_0[3]>", line 1, in <module>
> w = [x('factor(2^%s-1)'% randint(Integer(250),Integer(310)))
> for x in v]###line 23:
> sage: w = [x('factor(2^%s-1)'% randint(250,310)) for x in v]
> File "/Volumes/SageVolume/sage/sage-3.1.3.alpha0/local/lib/
> python2.5/site-packages/sage/interfaces/sage0.py", line 225, in
> __call__
> return SageElement(self, x)
> File "/Volumes/SageVolume/sage/sage-3.1.3.alpha0/local/lib/
> python2.5/site-packages/sage/interfaces/expect.py", line 1281, in
> __init__
> raise TypeError, x
> TypeError: Unable to start sage
Lack of memory space. Not much you can do. This does not mean that
enough virtual space is available, but that for example clisp cannot
be started twice in that setting on the same machine. clisp is a piece
of shit, i.e. on RHEL 5/Itanium it claims to have no memory available
to it even though 8GB are free and during configure its mmap interface
reports it to work. Oh well ....
<SNIP more of the same>
>
> sage -t -long devel/sage/sage/matrix/
> matrix_symbolic_dense.pyx**************************************************
> ********************
> File "/Volumes/SageVolume/sage/sage-3.1.3.alpha0/tmp/
> matrix_symbolic_dense.py", line 804:
> sage: f(1,2)
> Expected:
> [0 1]
> [2 z]
> Got:
> Maxima crashed -- automatically restarting.
> [0 1]
> [2 z]
clisp sucks. I have seen Maxima segfault sbcl due to out of memory
conditions for a 32 bit process on Solaris more than a couple times.
So in your case it burries itself deeply into swap before failing.
Consequently game over.
>
> sage -t -long devel/sage/sage/schemes/elliptic_curves/
> ell_finite_field.py
> halt 14
>
> error: no more memory
> System 5116k:5116k Appl 4668k/447k Malloc 4089k/2k Valloc 1024k/444k
> Pages 160/96 Regions 2:2
>
> A mysterious error (perphaps a memory error?) occurred, which may have
> crashed doctest.
> [205.0 s]
I think this is likely the pari precision issue, but we are dealing
with that.
> <SNIP>
>
> sage -t -long devel/sage/sage/schemes/elliptic_curves/
> ell_rational_field.py*** *** Error: TIMED OUT! *** ***
> *** *** Error: TIMED OUT! *** ***
> [3606.1 s]
>
> <SNIP>
>
> sage -t -long devel/sage/sage/schemes/elliptic_curves/sha.py*** ***
> Error: TIMED OUT! *** ***
> *** *** Error: TIMED OUT! *** ***
> [2719.1 s]
>
> As you can see, only the "ell_rational_field.py" doctest seems to have
> been hitting the 30 minutes == 3600s timeout barrier.
Sure
> The *very* long doctest above for "refinement_matrices.pyx" was really
> not 45000s, but eternity. More precisely, it really ran thus long, but
> then I killed manually one process. Only after that, the testsuite
> continued. This means, that in this special case your 30mins timeout
> timer did not do its job, the box would have run this one test
> endlessly.
Yes, the timeout code in Sage for the doctests is plain and simply
broken. Since I run into similar issues with endlessly spinning jobs
on occasion on Solaris I want this fixed. One way to do this is to set
ulimits to the shell that executes the individual doctests, but that
does not kill the "just sitting there due to pexpect synconization
issue" case, but would be better than no fix for now :)
> (This kind of endless running hits this PPC box randomly, last time
> e.g. in "desolver.py", which this time just passed OK.)
Well, most of those are swap related one way the other or due to clisp
screwing up. "'long" is not meant for every machine, that is what
"make check" is for.
> Cheers,
> gsw
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to [email protected]
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://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
