Initialize SimulatorAccess object for compositing material models. #673
Conversation
|
Let's give @Shangxin-Liu a chance to comment whether this works for him before merging. |
|
I've put the change to the public version of ASPECT on our cluster @bangerth. The compilation has passed. However, when I make the compositing material model test with the depth-dependent and simple compressible model, the segmentation fault still appears: [hs002:03064] *** Process received signal *** [hs002:03064] *** End of error message *** |
|
OK. I'll investigate tomorrow. |
|
Thanks. If you need my prm file to make the tests on your cluster, I can send you my depth-dependent viscosity file. |
|
Yes, either send it to me or copy it here. |
This fixes the fact that compositing material models (specifically the depth dependent and averaging material models) held pointers to base models that may need to access the simulator, but never initialized these base models' pointers to the Simulator object. This patch fixes this by intercepting the initializing call.
bangerth
force-pushed
the
fix-uninitialized-SimulatorAccess
branch
from
b9a0909
to
68ac161
Compare
| * a section added to the manual on how to use these parameters to | ||
| * <li> Fixed: Whenever the base models used by either the "depth dependent" | ||
| * or "averaging" material models depended on anything that requires accessing | ||
| * the simulator, then this led segmentation faults. This is now fixed. |
There was a problem hiding this comment.
Hi Wolfgang and Timo,
The segmentation fault problem has been solved. Thanks!
Shangxin
On Nov 25, 2015 9:14 PM, "Timo Heister" notifications@github.com wrote:
In doc/modules/changes.h
#673 (comment):@@ -6,9 +6,15 @@
*
*
- \*
- New: The tolerance of the preconditioners of the A and S block - \* are now available as parameters in the prm file. There is now also - \* a section added to the manual on how to use these parameters to - \*
- Fixed: Whenever the base models used by either the "depth dependent" - \* or "averaging" material models depended on anything that requires accessing - \* the simulator, then this led segmentation faults. This is now fixed.
"led to"?
—
Reply to this email directly or view it on GitHub
https://github.com/geodynamics/aspect/pull/673/files#r45938530.
|
looks good, please merge yourself after fixing the typo. |
Also fix a lot of lines with trailing whitespace.
bangerth
force-pushed
the
fix-uninitialized-SimulatorAccess
branch
from
68ac161
to
e2dfff0
Compare
Initialize SimulatorAccess object for compositing material models.
This fixes the fact that compositing material models (specifically the depth dependent
and averaging material models) held pointers to base models that may need to access
the simulator, but never initialized these base models' pointers to the Simulator
object. This patch fixes this by intercepting the initializing call.