WIP: Fix memory leak in ProtonPairsToDistanceDrivenProjection

[acoussat]

When running a Python script like

for p in range(number_of_projections):
    pct.pctbinning(input=os.path.join(pairs_folder, f'pairs{p:04d}.mhd'),
                   output=os.path.join(projections_folder, f'projections{p}.mhd'),
                   source=-1000.,
                   size=[200, 1, 200],
                   spacing=[2., 1., 1.],
                   verbose=True)

the memory consumption increases to a suspiciously high amount (dozens of gigabytes). After careful investigation using valgrind, it seems that the variable m_ConvFunc used in ProtonPairsToDistanceDrivenProjection is never freed, resulting in a leak of about 50MB per call to pct.pctbinning. These 50MB seem to correspond to the Bethe-Bloch LUT.

This wasn't a problem when the same script was ran through bash, because (to the best of my understanding) each pctbinning invocation resulted in a new sub-process that was deleted after completion. With Python, all the computations are executed in a single Python process, hence the issue.

I am not sure if my fix here is really "ITK-like", but as far as I tested this seems to do the trick. I assume that there is a way to fix it that better respects ITK's conventions.

[SimonRit]

Thanks (and sorry for the bug). I would move this to AfterThreadedGenerateData since it's been allocated in BeforeThreadedGenerateData.
We could also consider making this LUT a singleton to avoid recreating it if the parameters are not changed but that can be a future work.

[acoussat]

Thanks for the feedback. I moved the deletion to AfterThreadedGenerateData as you suggest in 0e37fec.

[acoussat]

The CI fails with the following error:

Imported target "RTK" includes non-existent path

     "/work/utilities/lp_solve"
 
   in its INTERFACE_INCLUDE_DIRECTORIES.  Possible reasons include:
 
   * The path was deleted, renamed, or moved to another location.
 
   * An install or uninstall procedure did not complete successfully.
 
   * The installation package was faulty and references files it does not
   provide.

The error seems to be linked with RTKConsortium/RTK#901. Do we need to change anything on PCT side? (@SimonRit @axel-grc)

[SimonRit]

I don't think it should be in this if (m_ComputeNoise) block. It can be deleted unconditionally I believe.

[acoussat]

Oops, my bad, I was sure I was at the end of the function when I added the line! Fixed in 00909ac.

[SimonRit]

The CI fails with the following error:

Imported target "RTK" includes non-existent path

     "/work/utilities/lp_solve"
 
   in its INTERFACE_INCLUDE_DIRECTORIES.  Possible reasons include:
 
   * The path was deleted, renamed, or moved to another location.
 
   * An install or uninstall procedure did not complete successfully.
 
   * The installation package was faulty and references files it does not
   provide.

The error seems to be linked with RTKConsortium/RTK#901. Do we need to change anything on PCT side? (@SimonRit @axel-grc)

I don't know what is the issue but it seems to be on the RTK side, I don't think you should have to change PCT. Maybe something's wrong in these lines. I don't know how to fix it though... Maybe @axel-grc can have a look?

[acoussat]

For future reference, we should check this PR again once RTKConsortium/RTK#918 is merged.