PI-1.8-1054 Computing View Properties hang - also how to recover ?

[bitli]

After a sequence like going back in history of an image, applying a clone stamp, double clicking an existing DBE icon and clicking somewhere else (just too fast, unfortunately I do not know where), the box 'Computing View Properties' showed up and never disappeared. There was no significant CPU used by PI, but no way to close the box. Had to kill and restart.
Sorry for the fuzzy sequence, but this was after a significant session where I lost work, I need to redo it first (and hope it wont happen again...).

By the way, there is an 'automatic workspace save' parameter. How do we recover the workspace after a crash ?
-- bitli

[StefanG]

This is interesting! I also managed to end up with this box 'Computing View Properties' by clicking somewhere to fast. In my case it was around image integration while I moved different windows around that did not respond quickly enough....
I had the 24bit STF activated (8GB Ram). Could that be causing this?
The possiblity to recover the workspace would have been handy, indeed!

Stefan

[bitli]

I'm glad that I am not alone. I do not think that I had 24 bit enabled. I had stf, it was a large RGB linear image.few other images open and large memory 12GB, windows 7. I likely pressed on the image juste after launching dbe (may be dbe tried to add another point), likely before the image was ready. The image was half off screen.
Hope this additional precision may help.
-- bitli

[Juan Conejero]

I have hardened the view property calculation routines with reentrancy barriers and making them more robust to spurious events. These problems should not happen anymore in the final 1.8.0 version.

There's no way to recover the state of the whole platform after a crash, unless you had saved a project previously. By default, the PSM AutoSave feature is enabled, so the process icon set is saved automatically every 30 seconds (this is configurable via Preferences) by a background task. The icons are saved as a binary .psm file on the system temporary directory (/tmp on UNIX/Linux). You can load this file after a crash.

[bitli]

Thanks, I was able to recover the icons and I start reprocessing from the last save of the workspace.
... CRASH
So it seems to be systematic. I was able to reproduce it by selecting a linear non-STFed image and then launching a DBE icon (the one that was saved automatically).
If I do an STF on the image before launching the DBE icon (and don't touch anything) it works fine. This was repeated twice, so there seems to be somehing related to DBE/STF too.
-- bitli

[kwiechen]

I have problems with build 1065, Win7 when launching a DBE icon. The 'Computing View Properties' box never closes and I have to restart PI.

Kai

[Juan Conejero]

If confirmed, this problem can delay the final 1.8.0 release. Could you upload the image and DBE icon that is causing this problem? That would be of great help to diagnose the problem, since as I am unable to reproduce it, it seems image-specific.

[Juan Conejero]

Kai,

Do you have any non-standard modules installed? Revising the code, I see that there's the possibility that if a module crashes or enters an infinite loop, the "Computing view properties..." dialog could never be hidden. This should never happen with the standard modules. I have modified the routine in question to prevent this problem.

[bitli]

I have not yet tested this aspect of the 1065. I will test tonight with the image that caused problem on the previous release (hopefully I still have it around) and keep you informed, unless the problem is solved until then....

I did test my scripts and some denoising process and workspace load with 1065 and it worked all fine.

-- bitli

[Juan Conejero]

OK, I've been able to reproduce it. It is a deadlock problem that only happens with the DBE tool. Fixing now ... thank you for your patience.

[kwiechen]

Juan,

test image with DBE icon: https://dl.dropboxusercontent.com/u/52792206/RGB.zip

Sorry for this late answer, but I did not have access to my data this day.

Kai

[Juan Conejero]

Hi Kai,

Thank you so much for uploading these files. Nice image BTW (but the gradients are hard to remove!). I have fixed this bug, so we have one less obstacle to release a final version...

[kwiechen]

Hi Juan,

this is 2xbin chrominance data collected with a 20" f9 cassegrain from 7th November to 9th of December. So there were varying additive gradients. It may be better to crop and DBE individual frames before integration??

Kai