<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Dear Dr. Geuzaine,<br>
<div align="justify">I think I may have found a bug in gmsh. The gmsh
binary compiled with gcc 3.4.3 under Mandrake Linux 10.2, with
optimization level 2 (-O2), is unable to build the 3D mesh for the
rect.geo file in attach. It starts the meshing, allocate more and more
memory until it crashes with the following error message:<br>
<pre>[gborzi@localhost gmsh-1.60.1]$ terminate called after throwing an instance of 'std::bad_alloc'</pre>
<pre> what(): St9bad_alloc</pre>
When compiled with optimization turned off (-O0), gmsh is able to do
the meshing, but again allocates a lot of memory (1GB RAM + ~2GB swap).
When the lc parameter is set to 0.15 or 0.5 both the optimized and
non-optimazed versions work OK.<br>
I understand that I may annoy you, but I have some other questions:<br>
<ol>
<li>When I tried to compile gmsh without optimization I tried <i>OPTIM=-O0
./configure</i> and also <i>./configure OPTIM=-O0</i>, but in each
case it uses -O2. I had to change the value in the variables file. How
should I give the command to get the right optimization ?</li>
<li>I'm developing a finite element code (it's still at version
2.0e-4 !) with gmsh as the pre and postpocessor. In order to check if a
2nd order simplex (triangle or tetrahedra) is curved or not my program
checks that the points <i>pm</i> in the middle of the edges are close
to the midpoint of the extremes of the edges <i>p1, p2</i>, i.e. <i>norm(pm-0.5*(p1+p2))/norm(p1-p2)
< epsnl</i>, where norm is the 1-norm. Is this an overshoot, i.e.
there is some better way to do the check ? For example by checking that
the average of the points at the vertices is close to the average of
the midpoints ? Also I guessed <i>epsnl = 1.0e-3</i>, but I'm not sure
that this a correct value. May you suggest me a criterion to check the
curved elements ?</li>
<li>When I make a 3D view of the results I do not see a solid figure,
but isosurfaces of the solution. Is this the standard behaviour of gmsh
or I did something wrong ?</li>
</ol>
Thank you in advance for your answers, and thanks also for developing
gmsh.<br>
<br>
Regards.<br>
</div>
<pre class="moz-signature" cols="72">--
Giuseppe Borzi'
e-mail: <a class="moz-txt-link-abbreviated" href="mailto:gborzi@ieee.org">gborzi@ieee.org</a>
<a class="moz-txt-link-freetext" href="http://www.webalice.it/gborzi1/">http://www.webalice.it/gborzi1/</a>
Assistant Professor at the Univ. of Messina - Italia
----------------------------------------------------
</pre>
</body>
</html>