[aspect-devel] Segmentation fault when referring to the functions of "class SimulatorAccess" in both depth-dependent.cc and its base model.

[Timo Heister]

Can you please post the code/prm that causes the error?
On Nov 23, 2015 23:33, "Shangxin Liu" <sxliu at vt.edu> wrote:

> Hi;
>
> I added some new stuff to the simple.cc file and this requires using "this
> -> get_geometry_model().depth(position)" to refer to the functions of the
> class SimulatorAccess. However, when I run the code with the compositing
> material model, i.e., a depth-dependent model with the base model of the
> simple material model, I always got a segmentation fault error as the
> follows:
>
> [hs002:21123] *** Process received signal ***
> [hs002:21123] Signal: Segmentation fault (11)
> [hs002:21123] Signal code: Address not mapped (1)
> [hs002:21123] Failing at address: 0xab0
> [hs002:21123] [ 0] /lib64/libpthread.so.0(+0xf500) [0x7fd24688e500]
> [hs002:21123] [ 1]
> /home/shangxin/aspect/build/aspect(_ZNK6aspect15SimulatorAccessILi3EE18get_geometry_modelEv+0x4)
> [0x81dcf4]
> [hs002:21123] [ 2]
> /home/shangxin/aspect/build/aspect(_ZNK6aspect13MaterialModel6SimpleILi3EE8evaluateERKNS0_19MaterialModelInputsILi3EEERNS0_20MaterialModelOutputsILi3EEE+0x99)
> [0x83ac49]
> [hs002:21123] [ 3]
> /home/shangxin/aspect/build/aspect(_ZNK6aspect13MaterialModel14DepthDependentILi3EE8evaluateERKNS0_19MaterialModelInputsILi3EEERNS0_20MaterialModelOutputsILi3EEE+0x27)
> [0x87cfa7]
> [hs002:21123] [ 4]
> /home/shangxin/aspect/build/aspect(_ZN6aspect19AdiabaticConditions14InitialProfileILi3EE10initializeEv+0x48e)
> [0x69759e]
> [hs002:21123] [ 5]
> /home/shangxin/aspect/build/aspect(_ZN6aspect9SimulatorILi3EEC2EP19ompi_communicator_tRN6dealii16ParameterHandlerE+0x1906)
> [0x7c56d6]
> [hs002:21123] [ 6] /home/shangxin/aspect/build/aspect(main+0x468)
> [0x693748]
> [hs002:21123] [ 7] /lib64/libc.so.6(__libc_start_main+0xfd)
> [0x7fd24650bcdd]
> [hs002:21123] [ 8] /home/shangxin/aspect/build/aspect() [0x68cd65]
> [hs002:21123] *** End of error message ***
> --------------------------------------------------------------------------
>
> At the beginning, I thought this was due to some problems of my new plugin
> in the simple.cc. So I went back to grab the original public ASPECT version
> from git hub and make a test using depth dependent model with the simple
> compressible as the base model. Unfortunately, this still leads to the
> segmentation fault as the follows:
>
> [hs002:31949] *** Process received signal ***
> [hs002:31949] Signal: Segmentation fault (11)
> [hs002:31949] Signal code: Address not mapped (1)
> [hs002:31949] Failing at address: 0x40
> [hs002:31949] [ 0] /lib64/libpthread.so.0(+0xf500) [0x7f7559add500]
> [hs002:31949] [ 1]
> /home/shangxin/aspect/build/aspect(_ZNK6aspect15SimulatorAccessILi3EE20get_surface_pressureEv+0x4)
> [0x81db04]
>
> Finally, I found that this error will always appear when using the
> compositing material model where the "class SimulatorAccess" is accessed
> both in depth-dependent and the base material model with the form such as
> "this -> get_geometry_model().depth(position)". And when I use the original
> simple.cc as the base model, which doesn't access the function of the
> "class SimulatorAccess" using command "this -> get_.......", the code can
> run successfully.
>
> It seems that the functions of the "class SimulatorAccess" cannot be
> accessed both in depth-dependent model and its base model using "this ->
> get_........".
>
> I think for now the expedient solution is to only use the simple.cc
> material model and add some new plugin to it for the depth-dependent data.
> But I'd like to report this problem with the compositing material model. If
> this can be fixed, it will be better to use the available
> "depth-dependent.cc".
>
> Best,
> Shangxin
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20151124/58e5fd1a/attachment.html>