It won't let me reply any more...
Just updating this thread to point out this can still happen. This
is on Cygwin; failed twice at the same spot, then no problems.
make[2]: Leaving directory
`/home/Administrator/sage-4.7.2/spkg/build/zn_poly-0.9.p5/src'
zn_poly tuning program
(use -v flag for verbose output)
Calibrating cycle counter... ok (2.37e+09)
mpn smp kara: done
mpn mulmid fallback: done
KS1/2/4 mul: ...............................
KS1/2/4 sqr: ...............................
KS1/2/4 mulmid: ...............................
nuss mul: ...............................
nuss sqr: ...............................
KS/FFT mul: ......................../spkg-install: line 59: 3460
Segmentation fault (core dumed) tune/tune > src/tuning.c
Error running tune program.
real 0m25.990s
user 0m18.081s
sys 0m1.781s
sage: An error occurred while installing zn_poly-0.9.p5
Please email sage-devel http://groups.google.com/group/sage-devel
---------- Forwarded message ----------
From: Bill Hart <goodwillh...@googlemail.com>
Date: Apr 5, 6:57 am
Subject: Unexpected failure on a test with znpoly
To: sage-devel
It is testing exceptionally simple code, with exceptionally simple
tests.
I doubt this is a bug in zn_poly.
If it turns out to not be reproducible, I would ignore it. It's much
more likely a cosmic ray or transient hardware fault or bug in the OS.
But let's see if it can be reproduced.
Bill.
On Apr 5, 1:28 am, "Dr. David Kirkby" <david.kir...@onetel.net> wrote:
> On 04/ 4/11 01:01 PM, Bill Hart wrote:
> > Is the problem reproducible?
> I'll have to check how reproducible it is by building znpoly repeatedly. But
> it's the first failure I've known in what must be more than 100 compete builds
> of Sage on this machine.
> znpoly is only taking 39 s to build and run the short test suite, so if I run
> it for a day, I can build+test it 2200 times. The more comprehensive test
> suite
> takes several minutes.
> But I've built Sage on this machine perhaps 100 times now, and the buildbot
> uses
> it too. So far only one failure.
> > If so, a valgrind log would be useful to debug the problem.
> Valgrind is not supported on this operating system.
> > It should
> > find the source of any segfault.
> > zn_poly would need to be compiled
> > with the -g option to GCC. It could of course be a compiler bug.
> > Possibly compilation with a lower optimisation level would make it
> > disappear.
> It could be. Whatever it is, it a rare failure, though just how rare I don't
> know now. The program is compiled with -O3, so plenty of scope for reducing
> that.
> The package compiles without any warnings at all, though the -Wall option is
> not
> added.
> Had parallel builds been enabled in the package, I might have put it down to a
> race condition, but they are not, so it can't be that.
> Dave
--
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
