Re: [Libmesh-devel] Bugfix in cell_tet10

[Roy S. <roy...@ic...>]

On Mon, 28 Feb 2011, Johannes Huber wrote:

> sorry, I tried to submit a patch file as descriped on the libMesh-Homepage,
> but it didn't work. What exactly do I have to type in? svn ci didn't do the
> job, I guess, I don't have the permission to add a file (which makes sense).

Right - we give anonymous svn read permission to everyone, but we only
give write permission to a few people, usually only after they've
already proven themselves by writing a lot of good libMesh code that's
been vetted by other already-proven developers.

> But I though the following patch is worth sharing it.

Of course, I mean "proven" in a loose, almost ironic sense, seeing as
how your patch fixes a long-standing horrible bug I wrote.

Thank you!

I changed your constants from "2." to "2", on the theory that
Real-divided-by-int should produce identical code to
Real-divided-by-double for all our current definitions of Real, but
the former might optimize better if we ever add a fixed-point
arithmetic option for representing geometry.

(Not that we worry enough about optimization - I think "x*.5"
evaluates faster than "x/2." with IEEE-conforming compilers, but we
pretty much always write the latter code for readability)

Anyway, Tet10::has_affine_map() should be fixed in svn now.  I'm
tempted not to bother releasing a 0.7.0.5 version with the bugfix,
since false negatives in has_affine_map() should only mess up
optimization and not cause any incorrect results...  But maybe I'm
wrong about that - how did you notice this bug in the first place?
---
Roy