Ticket #11038 (closed: fixed)
Random segfaults in doc tests
Reported by: | Martyn Gigg | Owned by: | Martyn Gigg |
---|---|---|---|
Priority: | critical | Milestone: | Release 3.4 |
Component: | Documentation | Keywords: | |
Cc: | Blocked By: | ||
Blocking: | Tester: | Dan Nixon |
Description
The documentation tests on RHEL6 have been unstable for some time. It seems to segfault when trying to load an instrument that doesn't exist:
Document: algorithms/CalculatePeaksHKL-v1 12:13:56 ----------------------------------------- 12:13:56 Load-[Notice] Load started 12:13:56 InstrumentDefinitionParser-[Warning] Attribute 'mark-as' is a depricated attribute in Instrument Definition File. Please see the deprecated section of www.mantidproject.org/IDF for how to remove this warning message 12:13:57 CreateSimulationWorkspace-[Error] Error in execution of algorithm CreateSimulationWorkspace: 12:13:57 CreateSimulationWorkspace-[Error] Unable to find an Instrument Definition File for search object MARS 12:13:57 MantidApplication-[Fatal] Unexpected exception: Unable to find an Instrument Definition File for search object MARS 12:13:57 CreateSimulationWorkspace-[Error] Error in execution of algorithm CreateSimulationWorkspace: 12:13:57 CreateSimulationWorkspace-[Error] Unable to find an Instrument Definition File for search object MARS 12:14:07 /var/tmp/sclwtzpCG: line 8: 24186 Segmentation fault (core dumped) /opt/Mantid/bin/MantidPlot_exe -xq docs/runsphinx_doctest.py
Change History
comment:2 Changed 6 years ago by Martyn Gigg
Remove setting/unsetting facility unecessarily
In MantidPlot, when the indirect interface is open, it causes a lot of flicking and random segfaults in the usage tests Refs #11038
Changeset: 423b4517e2f40dc36f56f7993d05ba2583ad5a3b
comment:5 Changed 6 years ago by Martyn Gigg
- Status changed from inprogress to verify
- Resolution set to fixed
This is being verified as pull request #188.
comment:6 Changed 6 years ago by Dan Nixon
- Status changed from verify to verifying
- Tester set to Dan Nixon
comment:7 Changed 6 years ago by Dan Nixon
- Status changed from verifying to closed
Merge pull request #188 from mantidproject/11038_fix_doctest_segfault
Remove setting/unsetting facility unnecessarily in indirect scripts
Full changeset: 18b5c92c5e1a45a4ad0ab0cf900a3fc7c23520c7
comment:8 Changed 6 years ago by Dan Nixon
For what it's worth this error also occurs within MantidPlot if you have multiple interfaces with instrument selectors on as by default they change the default facility.
IndirectInstrumentConfig helps to stop this to a certain extent by limiting the number of times the default instrument is changed.
comment:9 Changed 6 years ago by Martyn Gigg
Yes, I'm starting to think that changing the default.facility is not a good idea or at least it should be optional
comment:10 Changed 6 years ago by Dan Nixon
As part of #11048(http://trac.mantidproject.org/mantid/ticket/11048) I will add the option to have at least MWRunFIles not depend on the default instrument, so there should no longer be an issue with the indirect interfaces changing the default instrument.
InstrumentSelector at least has the updateInstrumentOnSelection() function to disable changing the default instrument.
comment:11 Changed 5 years ago by Stuart Campbell
This ticket has been transferred to github issue 11877
I put a breakpoint in the part of LoadInstrument that throws the exception and found this backtrace
My assumption is that a test is setting the default instrument and causing an InstrumentSelector widget to asynchronously update itself causing an occasional failure. The culprit seems to be IndirectInstrumentConfig