Crash when opening project with spectrogram

[Dan Nixon]

Mantid crashes when opening a saved workspace with a spectrogram plotted from a matrix.

To reproduce:

• Open Mantid and run the attached Python script
• Save the project and close Mantid
• Open Mantid and try to load the saved project

Here is a stack trace of the error

#5  0x00000000006e2deb in QList<QString>::operator[] (this=0x7fffffffbb90, i=1)
    at /usr/include/qt4/QtCore/qlist.h:477
#6  0x0000000000696335 in ApplicationWindow::openGraph (this=0x1c03000, 
    app=0x1c03000, plot=0x1d23a00, list=...)
    at /home/dan/mantid2/mantid/Code/Mantid/MantidPlot/src/ApplicationWindow.cpp:11907
#7  0x00000000006466a8 in ApplicationWindow::openProject (this=0x1c03000, 
    fn=..., factorySettings=false, newProject=false)
    at /home/dan/mantid2/mantid/Code/Mantid/MantidPlot/src/ApplicationWindow.cpp:4728
#8  0x0000000000643019 in ApplicationWindow::open (this=0x1c03000, fn=..., 

[Dan Nixon]

The code seems to be trying to load the workspace name from around here somewhere

<graph>
ggeometry	5	5	492	373
PlotTitle	ws2D	#000000	4228

Which makes sense for this plot which was generated from a workspace named ws2D, however this project also fails to load and produces an index out of bounds error at the same point.

[Dan Nixon]

(In #8503) I would only want to call this finished when I see if the code which appears to load plots from files is actually used.

[Harry Jeffery]

MantidPlot no longer crashes as a result of loading a spectrogram plot (color fill). v3.2 will always crash, and there is nothing we can do about that, and v3.2 will always save them incorrectly, meaning we cannot load spectograms from projects saved by v3.2. However, since #9970, Mantid does load and save spectrograms correctly, so this crash is fixed.

Compatibility with v3.2 is something that simply cannot be achieved though.

Testing

• Play with spectogram plots
  • Load/save them
  • Verify Mantid handles them as you would expect

[Dan Nixon]

Still not working, I get a data search exception now, it looks like it's looking for a workspace but the script creates it from a matrix (not MatrixWorkspace).

Also I get a segfault when I close the empty plot and reproduce #10679 (probably because there is no workspace).

[Harry Jeffery]

Refs #9952 Open plots after matrices

We can't plot a matrix if it's not been loaded yet.

Changeset: b3700ecfe0f580bbc3856405269293e8304b8411

[Harry Jeffery]

Refs #9952 Rework spectrogram source saving

Since v3.2 and prior crash when dealing with spectrograms, we don't have to worry about compatibility, so we're allowed to deal with this a little more simply.

If the spectrogram's based off a matrix, we specify the matrix by its name. If the spectrogram's based off a workspace, we specify the workspace by its name.

Load and save accordingly.

Changeset: d35220179d6acabb7377361d7f60907a53b24e7e

[Harry Jeffery]

Spectograms of matrices (not workspaces) are now loaded/saved correctly.

[Dan Nixon]

Merge remote-tracking branch 'origin/feature/9952_spectrogram_crash'

Full changeset: ​ce08a5c1371bc9a0e467f60b91d067fa4993cd78

[Stuart Campbell]

This ticket has been transferred to github issue ​10794