1.8RC7 osx - lockup / unresponsive

[pfile]

step 1: use DBE to create a background model.
step 2: save this background model as a .tif (f32)
step 3: use ImageCalibration to calibrate 3 frames (fits i16) using only this synthetic flat in f32 .tif

apply the calibration a couple of times with "overwrite existing files" checked and eventually the console gives:

*** Error
*** Error: Unknown exception

3 times, once for each input image.

after this 1.8RC7 is "locked up" in the sense that many menu options are greyed out and no UI elements can be clicked on.

this seems to be caused by using a .tif formatted image as the "flat" though i can't be sure.

under 1.7 there is no lockup but the output images are corrupted. when 1.8RC7 does not lock up, the output images are similarly corrupted.

attaching with gdb shows the backtrace as follows:

Code: [Select]

#0  0x00007fff9286339a in szone_size ()
#1  0x00007fff92861cec in realloc ()
#2  0x000000010397b932 in QListData::append ()
#3  0x00000001015f134c in qt_mac_execute_apple_script ()
#4  0x0000000101654ec5 in QApplication::topLevelWidgets ()
#5  0x00000001016111df in operator!= ()
#6  0x000000010006fc2f in pi::APIGlobalContext::GetProcessStatus ()
#7  0x000000011c4b8506 in pcl::Console::AbortRequested ()
#8  0x000000011c433f1c in pcl::ImageCalibrationInstance::ExecuteGlobal ()
#9  0x000000011c53cdb5 in pcl::ProcessContextDispatcher::ExecuteProcessGlobal ()
#10 0x0000000100afd217 in pi::MetaProcess::ExecuteGlobal ()
#11 0x0000000100b415cb in pi::ProcessInstance::ExecuteGlobal ()
#12 0x000000010006ddeb in pi::APIGlobalContext::LaunchProcessInstance ()
#13 0x000000011c54c847 in pcl::InterfaceDispatcher::ApplyInstanceGlobal ()

applying the DBE background model as a flat may not be a sane thing to do, but 1.8RC7 failing in this manner is really bad - i would have lost all my work (again) if i had been doing anything important.

i can upload a set of files if necessary.

rob

[Juan Conejero]

Hi Rob,

Yes, please upload a minimal set of files that reproduce this problem consistently. I'll try to fix it as soon as possible.

[pfile]

it was late last night, and looking at it this morning, now i know what the problem is. the background model has pixel dimensions which are 1/2 that of the original file.

i think IC is either not testing the image dimensions or the testing is somehow broken. it makes sense that the output is corrupted given the difference in pixel dimensions.

maybe that's enough to go on? i'll still upload the images.

rob

[pfile]

here is the set of files. probably tif vs. fit does not matter but i uploaded both versions.

if you try to calibrate the 3 images with the bogus flat, PI's UI will stay locked up. it may not happen the first time; could take a few applications of IC for it to happen. you'll see that the calibrated results are corrupted.

it's fine to have an undefined result here but doing something stupid should not lock up the UI...

thanks

rob

https://www.dropbox.com/s/qn60ox9czmo85zw/Calibration_mismatched_size.zip

[Juan Conejero]

Hi Rob,

Thanks for the images. I'm going to download them right now. Of course you're right; this should never happen. In theory ImageCalibration checks the geometry of all the images before starting to work, but this is not working well with these images for some reason. I'll investigate it.

[pfile]

any update on this? in a completely different thread you said you won't use dropbox so please let me know if you want me to upload these files again to some other service.

rob

[Juan Conejero]

Hi Rob,

I have just released an update that fixes this problem. Thank you for uploading the images.

Of course I can download public files from Dropbox. The problem is when the files are uploaded to a folder where one needs a Dropbox account to access.

[pfile]

thanks i'll apply the update later today, away from my PI computer right now...