[solved] thermal radiation related tests fail

Discussion about building and installing Elmer
Post Reply
KaiMartin
Posts: 34
Joined: 02 Jun 2013, 00:07
Antispam: Yes
Location: Hannover, Germany
Contact:

[solved] thermal radiation related tests fail

Post by KaiMartin » 22 Feb 2015, 22:43

Just did a compile from a fresh git clone on a Debian/jessie machine. Cmake configure and compile seemed to go smoothly without errors. This was my first exposure to cmake, I am very impressed by the build system.
However, some of the tests failed:

Code: Select all

97% tests passed, 10 tests failed out of 304
Total Test time (real) = 972.90 sec
The following tests FAILED:
	 27 - ParticleAdvectorPathIntegral (Failed)
	 41 - radiation2 (Failed)
	144 - ParticleHeating (Failed)
	149 - mgdyn_harmonic_loss (Failed)
	151 - radiation2d (Failed)
	171 - InductionHeating2 (Failed)
	179 - ThermalActuator (Failed)
	197 - radiation3d (Failed)
	248 - radiation (Failed)
	301 - mhd2 (Failed)
Errors while running CTest
Where would I look for the cause? Is there some log somewhere, which I can search for suspicious messages?
Most of the failed tests seem to related to thermal radiation. Would that be a lead?

Any hints appreciated.
Last edited by KaiMartin on 26 Feb 2015, 16:22, edited 1 time in total.

raback
Site Admin
Posts: 3450
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: thermal radiation related tests fail

Post by raback » 23 Feb 2015, 00:06

Hi Kai

There are currently 4 to 5 tests that fail because the norm has changed a little. We assume that this is nothing to worry about. If you are far away from the "Target Norm" in the test then there might be some more severe problem.

You do have a few extra tests that are usually passed. These seem indeed to be related to radiation. You could try to go to the tests directory and run them directly. I suspect that you might have some problem with the "ViewFactors" binary that is called by ElmerSolver. You can also run it directly and see if it performs without problems.

-Peter

annier
Posts: 1115
Joined: 27 Aug 2013, 13:51
Antispam: Yes

Re: thermal radiation related tests fail

Post by annier » 23 Feb 2015, 19:11

Hi Kai and Peter,
Even in 2 computers using the same operating system(OS), the tests that fail after installing Elmer via cmake, are different. You can see them at:
viewtopic.php?f=2&t=3742&sid=69538bf5db ... aed195d728

Can you guess what might be the reason for that?

Yours
Anil Kunwar
Anil Kunwar
Department of Materials Engineering, KU Leuven, Belgium

KaiMartin
Posts: 34
Joined: 02 Jun 2013, 00:07
Antispam: Yes
Location: Hannover, Germany
Contact:

Re: thermal radiation related tests fail

Post by KaiMartin » 24 Feb 2015, 03:58

Hi Peter,
raback wrote: You do have a few extra tests that are usually passed. These seem indeed to be related to radiation. You could try to go to the tests directory and run them directly. I suspect that you might have some problem with the "ViewFactors" binary that is called by ElmerSolver. You can also run it directly and see if it performs without problems.
Hmm. How do I run the tests manually?

Anyway, your suspicions seem to be spot-on. The file /usr/local/src/elmer/build/fem/tests/radiation2/test-stdout.log contains:

Code: Select all

ViewFactors:  E L M E R  V I E W F A C T O R S,  W E L C O M E
ViewFactors: ==================================================
ERROR:: systemc: Command exit status was 2
ERROR:: RadiationFactors: Memory allocation error 3.
/usr/local/src/elmer/build/fem/tests/radiation2/test-stderr.log says:

Code: Select all

TEST CASE 1
TEST CASE 1
At line 610 of file ElementDescription.f90 (unit = 1, file = '���	�')
Fortran runtime error: File '/usr/local/src/elmer/build/fem/src/modules/elements.def' does not exist
(broken characters in the original, too)

The log files for the ParticleHeating test look different. In this case, the test run completed but resulted in "Relative Error to reference: 1.608505E-02"

The log files for radiation2d are similar to the logs of radiation2. The test-stderr.log contains an additional backtrace:

Code: Select all

TEST CASE 1
TEST CASE 1
At line 610 of file ElementDescription.f90 (unit = 1, file = '����')
Fortran runtime error: File '/usr/local/src/elmer/build/fem/src/modules/elements.def' does not exist

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0  0x7F3845FB9407
#1  0x7F3845FB9A1E
#2  0x7F3844E3517F
#3  0x7F384681FF2A
#4  0x7F38448D6B75
#5  0x7F38448DA4FE
#6  0x7F3846880254
#7  0x7F384688FB5E
#8  0x7F3846890BE4
#9  0x7F3846893568
#10  0x7F38469FFD0A
#11  0x7F3846A0486C
#12  0x400F2E in solver at Solver.F90:69
The log files for InductionHeating2, radiation3d and radiation look the same -- they even report the exact same backtrace.

The project I'd like to apply elmer to does not involve thermal radiation. It is all about charged particles in a static electromagnetic field. So I decided to try to move on with the build. However, I ran into compile errors if I activate elmer-gui in the cmake config:

Code: Select all

/usr/local/src/elmer/build$ make -j4
[  0%] Built target amdf77
[  0%] Built target umfpack_srcs
[  4%] Built target amd
[  6%] Built target matc
(...snip...)
 89%] [ 90%] Built target ParticleAdvector
Built target ParticleDynamics
Generating moc_summaryeditor.cpp
[ 90%] Built target PhaseChangeSolve
Generating moc_curveeditor.cpp
[ 90%] Built target PoissonBEM
Generating moc_renderarea.cpp
[ 90%] [ 90%] Built target PoissonBoltzmannSolve
Built target PressurePrecond
[ 90%] Generating moc_twodview.cpp
Built target PressureUpdate
[ 90%] Built target RateOfChange
[ 90%] uic failed...
Built target ProjectToPlane
ElmerGUI/Application/CMakeFiles/ElmerGUI_automoc.dir/build.make:49: recipe for target 'ElmerGUI/Application/CMakeFiles/ElmerGUI_automoc' failed
make[2]: *** [ElmerGUI/Application/CMakeFiles/ElmerGUI_automoc] Error 1
CMakeFiles/Makefile2:13564: recipe for target 'ElmerGUI/Application/CMakeFiles/ElmerGUI_automoc.dir/all' failed
make[1]: *** [ElmerGUI/Application/CMakeFiles/ElmerGUI_automoc.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 90%] [ 91%] Built target ReloadData
Built target ReloadInput
[ 91%] Built target ResultOutputSolve
Makefile:147: recipe for target 'all' failed
make: *** [all] Error 2
Could whatever causes my radiation test problems possibly also affect the build of elmer-gui? Or is this a completely unrelated issue?

From my first attempts with elmer about three years ago I remember that open cascade is a critical ingredient for elmer-gui. I have about every open cascade related package installed the Debian/jessie repo currently offers. Are these known to work with elmer? Or should I go and build OCE from source too?
---<)kaimartin(>---

KaiMartin
Posts: 34
Joined: 02 Jun 2013, 00:07
Antispam: Yes
Location: Hannover, Germany
Contact:

Re: thermal radiation related tests fail

Post by KaiMartin » 26 Feb 2015, 03:20

Today, I tried a fresh git_clone and compile on a different linux desktop. This machine was set-up fairly recently. So there is little probability for fluff from my previous elmer attempts which might interfere with the current build. A compile without elmer-gui and without MPI went through without errors.

This time, only five tests failed.

Code: Select all

The following tests FAILED:
	  1 - ParticleHeating (Failed)
	 95 - ParticleAdvectorPathIntegral (Failed)
	205 - mgdyn_harmonic_loss (Failed)
	264 - ThermalActuator (Failed)
	294 - mhd2 (Failed)
Errors while running CTest

All of them failed the norm. None of them failed to run. These seem to be the harmless cases mentioned by Peter. So I guess, my problems on the older desktop were indeed triggered by some kind of fluff from the past. Problem resolved. :-)

However, a compile with elmer-gui still fails like reported in my previous post. Since this seems to be an independent issue, I will open a new thread for it.
---<)kaimartin(>---

KaiMartin
Posts: 34
Joined: 02 Jun 2013, 00:07
Antispam: Yes
Location: Hannover, Germany
Contact:

ui issue solved, too

Post by KaiMartin » 28 Feb 2015, 07:03

KaiMartin wrote: However, a compile with elmer-gui still fails like reported in my previous post.
For the archive: I was able to solve this issue, too. The compile looked for *.ui files in the wrong direction. Soft links to the actual files did the trick.
---<)kaimartin(>---

Post Reply