Crashing using the newly updated Photometric Mosaic script

B

Barry-Wilson

Active member
Opening PI today I installed the updated Photometric Mosaic script 3.2 and I am having repeated crashes using an iMac Catalina 10.15.7. The prior version of the script worked without issue.

The script opens fine and then when executing the first two panel joining of a 6 panel mosaic PI crashes and immediately closes the whole program. I have repeated this five times with a system re-start between. On two occasions, the script successfully merged the first two panels and then crashed merging the next pair with the same immediate closure of PI.

Otherwise PI is normally very stable for me on this Mac for all processes and scripts.

Is there anything I can do to help diagnosis?

Barry
 
I have just tried once more this time only opening two of the registered panels. Same experience of a crash and an immediate closure of PI. Here is a screenshot of the top part of the Apple bug report, if it helps.

1616268932670.png
 
can you post the two panels? that way at least @jmurphy can see if it can be replicated (if he has time)...

at first glance it does look like a crash in the javascript engine.

rob
 
Barry-Wilson said:
@jmurphy dropbox link to the 6 registered G subs, https://www.dropbox.com/sh/occmghlk4q6bo11/AAD3T9pPzWHR8dWWeVDuRrSHa?dl=0.

The blue channel mosaic was compiled last week with the prior version and it worked excellently (as it has on many other mosaics I have processed).

Thanks for your support and please let me know if there is anything further I can do to help diagnose.

Barry
Click to expand...
Hi Barry
I can't get it to crash on my Windows 10 machine, but when creating a mosaic from tiles 4 and 6, the surface spline produced is extreme. I wonder if this extreme surface spline is causing the problem to occur on your machine. For these two images, I get the same extreme surface spline on both version 3.0 and 3.2

To avoid getting this extreme surface spline, you need to trim more off the edges of your data files:
1616277196422.png

In the image above, you can see the the image outline (drawn in white) does not follow your image boundary. This indicates 'almost black' regions. The same area in image4 has a normal background value. The surface spline tries to model this - creating a very extreme result.

To test to see if this is the cause of your crash, could you trim the tiles back to good data and try again. The image above shows the trim values necessary for image6. You will need to determine the values necessary for your other tiles.
Let me know how you get on,
John Murphy
 
Last edited:
Thank you for your guidance John.

Trimming more stacking artefacts from the edges has been successful and the G channel is now a complete mosaic, saved to join the Blue and Ha. Great. Success! Red to go now . . .

Would you expect PI to crash and close when encountering this user input error or some other behaviour? Regardless, I now better understand the trimming function.

Many thanks once again.

Barry
 
Last edited:
hmm, no crash on 4+6. barry, was there a particular set of panes that caused the problem for you?

rob
 
Yes panels 1 and 2. I was proceeding with 1+2, then 3+4, than 5+6 to give three horizontal 1 x2 part-mosaics, before merging each 1 x 2 to give 2 x 2, then 2 x 3. I had multiple PI crashes before additional edge trimming.

John contacted me via PM with a series of tests to carry out. I had mixed results on the existing G panels with panels 1 and 2 - initial crashes of PI and then when re-opening PI, success. I then closed PI and after waiting a few minutes repeated the test and got the same results, an initial PI crash and then quickly re-opening PI I could successfully open the dialogue graphs and could merge the panels. Very odd.

I then re-integrated the aligned G subs and using just the two panels, 1 and 2, registered these and had success without any trimming at all. In this test the two panels were just registered to each other and not as part of an overall 6 panel mosaic.

I am now proceeding with my Red subs and will report back to John.

Barry
 
Last edited:
thanks - earlier this morning i was able to replicate the crash by asking for the overlap gradient with 4+6 selected. maybe @Juan Conejero can take a look at this on his mac development platform. maybe it is a bug in spidermonkey on OSX but even so perhaps there's a way to mitigate it in PI.

rob
 
Hi Rob

I have been providing update reports to John as I have been progressing with the mosaic preparation and merging, with similar progress interspersed with crashes and multiple PI re-starts and Mac reboots. This experience with the Red and Green channels is in complete contrast with the Ha and Blue channels using the prior version in which I had no issues whatsoever.

I now have each channel merged to a mosaic . . . so onwards with the processing ;-)

Hope this helps to diagnose if there is an issue.

Barry
 
John

I have re-booted the Mac and have bee able to merge the three partial mosaics into one complete mosaic without issue.

I have just updated Rob on the forum. This experience with the Red and Green channel mosaic prep and merging is in complete contrast to the Ha and Blue channel with the previous version one week ago in which there was no issue at all.

I realise this in itself doesn't help, but my experience separated by a week (or so) and the two versions is 'chalk & cheese'.

Anyway - I am happy to continue to test to help.

Best regards,

Barry
 
Rob

I have tested this afternoon following some direction from John, including installing ver 3.

Using ver 3 I was able to merge two panels which caused 3.2 to crash. John is now considering possible causes.

Barry
 
Hi @Barry-Wilson, @pfile
From the console image I received from Rob, I believe that the error is occurring within my photometry code. This section had become more complicated than it needed to be, and I had already planned to rewrite it. However, this rewrite did not make the 3.2 release. Since the error is probably in this section, I have now rewritten it. Not only has this simplified the code, it is now also more efficient.

I was unable to reproduce the error on Windows 10, so to some extent I am working blind. Please try this new 3.2.1 Beta version, especially if you use a Mac, and let me know if it fixes the problem.

To install:
(1) Unzip and save anywhere.
(2) SCRIPT -> Feature Scripts...
(3) Add (navigate to where you saved it)
(4) Done
It will install in the SCRIPT -> Mosaic menu as 'PhotometricMosaic 3.2.1'

Thanks
John Murphy
 
Hi @Barry-Wilson, @pfile

I have attached PhotometricMosaic V3.2.1 Beta2

Either uninstall Beta1, or copy the source code over the top of it. If you are running the correct version, the console will display:
=== Photometric Mosaic 3.2.1 Beta2 ===

This version prevents the StarDetection routine from attempting to read beyond the edge of the image. I am hoping that this will fix the remaining problems.

Thanks, John
 
while i didn't see the lockup, unfortunately i got a crash after playing around with the script a bit. i will try to find out what it was doing at the time of the crash.

rob
 
You must log in or register to reply here.
Back
Top