"Error: Unknown exception" persists with ImageIntegration build 445 on Mac

[johnpane]

* Integrating channel 1 of 3:
*** Error: Unknown exception
<* failed *>

There is a larger excerpt from the console log below. Files and ImageIntegration settings to reproduce are in this zip file on DropBox:
     https://www.dropbox.com/s/iob9tcy2r8k9pz4/PixInsight%20ImageIntegration%20bug%20report.zip

Weights is currently set to use a FITS keyword. Error does not occur if I change weights to "Noise evaluation". However, it still fails if I change weights to "Don't care."

Alternatively, turning off "Generate drizzle data" will avoid the error.

This was originally reported in https://pixinsight.com/forum/index.php?topic=11554.0, but the title on that thread was not directly applicable to this bug, so I am starting this new thread.

--- excerpt from console log ---
ImageIntegration: Global context

* Loaded cache: 420 item(s)

Opening files:

/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4954_c_d_a_r.xisf
/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4954_c_d_a_r.xnml
/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4954_c_d_a_r.xdrz
Loading image: w=5496 h=3670 n=3 RGB Float32
* Retrieved data from file cache.
45 FITS keyword(s) extracted.
Weight          :     1.00000    1.00000    1.00000

* Using 2 worker threads.
[000] /Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4955_c_d_a_r.xisf
[001] /Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4956_c_d_a_r.xisf

* Waiting for running tasks to terminate...

/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4955_c_d_a_r.xisf
/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4955_c_d_a_r.xnml
/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4955_c_d_a_r.xdrz
Loading image: w=5496 h=3670 n=3 RGB Float32
* Retrieved data from file cache.
45 FITS keyword(s) extracted.
Weight          :     0.99230    0.99230    0.99230

/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4956_c_d_a_r.xisf
/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4956_c_d_a_r.xnml
/Users/pane/Desktop/PixInsight ImageIntegration bug report/JFP_4956_c_d_a_r.xdrz
Loading image: w=5496 h=3670 n=3 RGB Float32
* Retrieved data from file cache.
45 FITS keyword(s) extracted.
Weight          :     0.98153    0.98153    0.98153


Integration of 3 images:
Pixel combination .................. average
Output normalization ............... local
Weighting mode ..................... custom keyword: INTEGRATIONWEIGHT
Scale estimator .................... iterative k-sigma / BWMV
Pixel rejection .................... none

* Using 763 concurrent pixel stack(s) = 159.97 MiB

* Integrating channel 1 of 3:
*** Error: Unknown exception
<* failed *>

[johnpane]

Still hoping this bug report will get some attention before it falls off the radar...

[Juan Conejero]

Hi John,

I still haven't been able to analyze this problem because I am too busy with other priority tasks currently (including a new RAW format module to overcome the end of dcraw support). I'll try to fix this problem as soon as possible; be sure it is not "off the radar" by any means.

[johnpane]

Thank you Juan. I certainly understand that you have to set priorities.

It is a bit surprising you aren't getting more reports of this problem. I have been unable to complete processing any of my datasets using both drizzle integration and a weights keyword written by SubframeSelector.

[johnpane]

Not sure if this will help you figure it out, but in ImageIntegrationInstance.cpp, all of the weight modes that fail (i.e. crash) assign a scalar value to the DVector m_weight, but the weight mode that succeeds (NoiseEvaluation) assigns the weight(s) to m_weight[c]. I don't know enough about C++ to know if assigning a scalar value to a vector could be the issue but it caught my eye.

[johnpane]

I now see that DVector is designed to do the right thing when a scalar is assigned to the vector.

The crash is occurring on line 4184 of ImageIntegration.cpp:
               s[i] = IntegrationFile::FileByIndex( i ).Scale( c );

In my setup, needScale is false and m_scale is thus unallocated when it is accessed in Scale().

[Juan Conejero]

Hi John,

This bug is now fixed with an update I've just released for all platforms. Your analysis has been of great help to find and solve the problem, thank you so much for that.

Also thank you for your patience with the long delay in fixing this problem; some family issues have kept me out of my normal work for a few weeks. Please let me know if the update fixed this bug completely.

[johnpane]

Thank you very much Juan. I understand family circumstances can take time from work - has happened to me - hope things have improved for your situation.