[petsc-dev] Fwd: segv with triangle

[Stefano Zampini]

Satish,

I have reported this issue to Matt some time ago but he probably forgot to
fix it.

Attached is new tarball for Triangle that fixes the issues with the Intel
compilers in next-tmp here
http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2018/11/21/examples_next-tmp_arch-pardiso_grind.log

The changes are minimal, here is the diff

diff -r Triangle/src/triangle.c TriangleOld/src/triangle.c
326,327d325
< #else /* Stefano Zampini - Thu Nov 22 10:35:29 +03 2018: we need to
include stdlib.h for malloc */
<   #include <stdlib.h>

Could you please upload it?

Thanks
Stefano

---------- Forwarded message ---------
From: Stefano Zampini <stefano.zampini at gmail.com>
Date: lun 15 mag 2017 alle ore 00:23
Subject: Re: segv with triangle
To: Matthew Knepley <knepley at gmail.com>


Matt, I did some test, and it seems related with your NO_PETSC_MALLOC define
you should include <stdlib.h> if NO_PETSC_MALLOC is not defined.
Adding the header file fixed the problem


On May 14, 2017, at 9:24 PM, Matthew Knepley <knepley at gmail.com> wrote:

On Sun, May 14, 2017 at 10:58 AM, Stefano Zampini <stefano.zampini at gmail.com
> wrote:

> And using the latest intel compiler. Could it be causing the problem?
>

Possibly. I would

a) Do a build with gcc. Last time this came up, the gcc build worked. The
triangle source is huge, and
    some compilers just shit the bed with their optimizations.

b) If that works, we might back off some optimizations

c) If not, we can get a valgrind trace, but I have never personally seen a
triangle bug

  Thanks,

     Matt

> On May 14, 2017, at 5:44 PM, Stefano Zampini <stefano.zampini at gmail.com>
> wrote:
>
> Forgot to say that it is a brand new installation I did, using the latest
> next
>
> On May 14, 2017, at 5:36 PM, Stefano Zampini <stefano.zampini at gmail.com>
> wrote:
>
> Matt,
>
> this is from next, on a machine I have never run on
>
> #0  0x00007ffff74735a1 in poolinit (pool=0x7ffff7dd9b80 <triangles>,
> bytecount=72, itemcount=4092, wtype=POINTER, alignment=4) at
> src/triangle.c:3227
> #1  0x00007ffff7473fe0 in initializetrisegpools () at src/triangle.c:3597
> #2  0x00007ffff748e5e6 in delaunay () at src/triangle.c:8648
> #3  0x00007ffff749870c in triangulate (triswitches=0x7fffffffb020 "pqezQ",
> in=0x7fffffffb1f0, out=0x7fffffffb298, vorout=0x0) at src/triangle.c:12993
> #4  0x00007ffff5e2ed22 in DMPlexGenerate_Triangle (boundary=0x8218f0,
> interpolate=PETSC_FALSE, dm=0x7fffffffc580)
>     at
> /marconi/home/userexternal/lpavarin/petsc/src/dm/impls/plex/plexgenerate.c:269
> #5  0x00007ffff5e3830c in DMPlexGenerate (boundary=0x8218f0, name=0x0,
> interpolate=PETSC_FALSE, mesh=0x7fffffffc580)
>     at
> /marconi/home/userexternal/lpavarin/petsc/src/dm/impls/plex/plexgenerate.c:1097
> #6  0x00007ffff5b50da4 in DMPlexCreateBoxMesh (comm=1140850688, dim=2,
> numFaces=2, interpolate=PETSC_FALSE, dm=0x7fffffffc580)
>     at
> /marconi/home/userexternal/lpavarin/petsc/src/dm/impls/plex/plexcreate.c:491
> #7  0x0000000000404ffa in CreateMesh (comm=1140850688,
> user=0x7fffffffc538, dm=0x7fffffffc580)
>     at
> /marconi/home/userexternal/lpavarin/petsc/src/snes/examples/tutorials/ex62.c:305
> #8  0x0000000000408b95 in main (argc=1, argv=0x7fffffffcba8) at
> /marconi/home/userexternal/lpavarin/petsc/src/snes/examples/tutorials/ex62.c:485
>
>
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener




-- 
Stefano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20181122/e25ad084/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Triangle_new.tar.gz
Type: application/gzip
Size: 143035 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20181122/e25ad084/attachment-0001.bin>