OK, I've got it. PR forthcoming. Manav was on the right track moving the
side_fe_reinit around. reinit needs to be get called to properly initialize
the quadrature rule. However, the reason this didn't fail in 2D or on the
edge in 3D was because QGauss (the default quadrature rule here) internally
No problem! This is still good info - probably means this has been broken since
I originally wrote that code.
On Sep 21, 2013, at 7:49 AM, Manav Bhatia wrote:
> Hi Paul,
>
> Now that I think of it, I had never before tested this for 3d cases.
>
> Sorry that I made the wrong claim
Hi Paul,
Now that I think of it, I had never before tested this for 3d cases.
Sorry that I made the wrong claim.
Manav
> On Sep 21, 2013, at 2:23 AM, "Paul T. Bauman" wrote:
>
> On Fri, Sep 20, 2013 at 10:26 PM, Paul T. Bauman wrote:
>>
>>
>> Thanks for the test case! I'll pl
On Fri, Sep 20, 2013 at 10:26 PM, Paul T. Bauman wrote:
>
> Thanks for the test case! I'll play with this in a little while and see if
> I can reproduce.
>
OK, I can reproduce. Will figure out the problem and get a PR in. You said
this worked previously - any chance you have a hash or version nu
Hi Manav,
On Fri, Sep 20, 2013 at 10:22 PM, Manav Bhatia wrote:
>Did you try this in 2D or 3D?
>
2D
> I do not get this error in 2D, but it shows up in 3D.
>
Hmm, OK.
>
>The following code is able to reproduce the issue at my end.
>
> int main (int argc, char* const argv[])
> {
>
Hi Paul,
Did you try this in 2D or 3D? I do not get this error in 2D, but it shows up
in 3D.
The following code is able to reproduce the issue at my end.
int main (int argc, char* const argv[])
{
// Initialize libMesh.
LibMeshInit init (argc, argv);
{
ParallelMe
Yes, it is 2239. I am sorry about the typo and the resulting confusion.
-Manav
On Fri, Sep 20, 2013 at 4:07 PM, Roy Stogner wrote:
>
> On Fri, 20 Sep 2013, Manav Bhatia wrote:
>
> Sorry I was not too specific. I am on the git master, and line 2339 is a
>> call to qsiderule.n_points(). It has an
Hmm. I can't reproduce this. I have a code that uses FEMFunction for
projections and I bumped up the basis order and cannot trip the error you
are. In fact, I manually verified that n_points() returns the correct value
on that line. Can you make a test case that trips the error? Are you using
a LAG
Just found the the call to side_fe_reinit() in this for loop. It seems like
the entire block from lines 2238 to 2247 should be moved to after the
context.side_fe_reinit() ?
-Manav
On Fri, Sep 20, 2013 at 4:21 PM, Manav Bhatia wrote:
> I was comparing the code with AssemblyContributions in fem_s
I was comparing the code with AssemblyContributions in fem_system.C, and
see that context.side_fe_reinit() has not been called before the call to
get qsiderule.n_points(). Should that be added to the for loop on element
sides at line 2248 and move the get_side_qrule method call (line
2238) after t
On Fri, 20 Sep 2013, Manav Bhatia wrote:
> Sorry I was not too specific. I am on the git master, and line 2339 is a
> call to qsiderule.n_points(). It has an assertion of
> libmesh_assert(!_points.empyt()), where the code throws an exception.
I think Paul's (and my) confusion was a simple typo -
Hi,
Have there been any changes to System::project_solution() with
FEMFunctionBase over the last 2-3 months?
I had a piece of code that was working fine with it, but now is throwing
an error at line 2339 in sytem_projection.C:
const unsigned int n_qp = qsiderule.n_points();
It seems
On Fri, Sep 20, 2013 at 2:45 PM, Paul T. Bauman wrote:
> On Fri, Sep 20, 2013 at 1:34 PM, Manav Bhatia wrote:
>
>>
>>Have there been any changes to System::project_solution() with
>> FEMFunctionBase over the last 2-3 months?
>>
>
> Current master shows FEMFunctionBase was touched 5 months ag
On Fri, Sep 20, 2013 at 1:34 PM, Manav Bhatia wrote:
>
>Have there been any changes to System::project_solution() with
> FEMFunctionBase over the last 2-3 months?
>
Current master shows FEMFunctionBase was touched 5 months ago.
>I had a piece of code that was working fine with it, but
14 matches
Mail list logo
