Reproduceable crash in star alignment on mac and linux

etill

New member
Mar 31, 2020
4
0
Hi,

Using star alignment to create a mosaic (register/union - separate) causes PI to crash on both linux (fedora, kde plasma) and Mac. Same images cause the issue on both platforms, two panels of the mosaic complete, two cause the crash. Using a rough mosaic to do the alignement which is quite a large image (~2.4gb). I have tried to reduce size saving as a 16bit image and also resampling down to smaller size just to test. Both cases still crash.

The attached image shows settings used in both cases, all other settings are deault. Crash report from Mac attached. I can probably get one from linux with gdb if that would help.

The Mac is a 15 inch 2017 model MBP, i7, 16gb ram, radeon graphics.
The linux machine is an HP proliant ML350, 64gb ram, dual xeon, quadro graphics.
 

Attachments

pfile

PTeam Member
Nov 23, 2009
5,321
64
more helpful to juan would be the images that cause the crash. can you upload them to a file hosting service and post the link here?

rob
 

etill

New member
Mar 31, 2020
4
0
If you know of one where I can upload 4 gig of images for free I can certainly do that. Do you have any you can recommend?

I thought I may wait and see if the crash report showed anything that is known first.
 

pfile

PTeam Member
Nov 23, 2009
5,321
64
you'll have to whittle it down to the minimum set that causes the problem. binary partitioning should help.

in the history of this project i'm not sure that the crashdump has ever been enough for juan to fix a problem.

rob
 

etill

New member
Mar 31, 2020
4
0
These would be the smallest option size wise that are triggering it on both platforms. I'm happy to upload them though if you have a support server or some such place to put them?

799M 31 Mar 18:26 drizzle_integration.xisf
2.5G 31 Mar 18:15 p4_mosaic.xisf
 

Juan Conejero

PTeam Member
Sep 2, 2004
7,403
67
56
Valencia, Spain
pixinsight.com
Well, in this case I think that the crash report text is sufficient to get an idea of what happens. The error is arising from calculation of recursive subsplines. This is probably a bug, especially because you are getting it on Linux, which is our reference platform.

I'm going to investigate this. At least two panels of the mosaic where this can be reproduced would be extremely useful, but if you can't upload them, don't worry. Anyway, I am going to replace the entire recursive subspline implementation with a much more efficient algorithm, which will fix these problems for sure.

As a workaroud, disable the local distortion option of StarAlignment just for this mosaic. This will prevent the use of recursive subsplines, and hence prevent the crash.

Sorry for the inconvenience.