Closed rozyczko closed 3 years ago
I have run multiple continuous fits with all engines, never getting a crash. Windows 10, develop branch
This bug seems not reproducible or, let's say, half reproducible on MacOS 10.13.6 (easyAppGui
: develop
branch, easyDiffractionApp
: fit_stop
branch). I launch easyDiffraction
via Poetry
/PyCharm
, load the PbSO4 project, go to the analysis page, select 1 parameter to refine (lattice constant a) and click the start fitting button. Approximately 50% of the time the fitting is finished properly, but in the other 50% of the time the application crashes with a message: Process finished with exit code 139 (interrupted by signal 11: SIGSEGV).
Windows 10, easyAppGui
and easyDiffractionApp
branches the same as yours, launching from both poetry run easyDiffraction
and from within visual studio code
, load PbSO4
and performed the same set of fits.
None of them failed. Changed parameters, switched to another Windows machine with the same result.
Reminds me the previous issue of windows start failures which would not show up on my dev machines but only on a clean setup.
I'll try next on a clean Windows VM, installing the pre-release package
Same on a clean Windows VM - no crashes with the most recent row_scroll
branch installer
Seems therefore to be a MacOS specific issue
Tested on bare metal mac mini (192.168.1.40) on installed version from the develop
branch.
As noticed by @AndrewSazonov , the app fails every other/third time fitting is run.
This means it is a mac os specific issue.
First thing to check if threading. I'll create a build based on the develop
source but without threading and test it on the VM
The problem does seem to originate with threading after all. The non-threaded version of develop
works just fine on the test mac.
As mentioned, everything works on Windows, but occasionally, when closing the app I see
QThread: Destroyed while thread is still running
which means some threads haven't been properly terminated.
This seems to be benign on Windows but might cause issues (as it should!) on MacOS.
Most recent commit in fit_stop
seems to have fixed the issue.
Please test the installer from that branch.
fit_stop
also contains code/functionality for stopping the fitting thread - it would be extremely useful to have this in the release.
BUG