another crash :( osx 10.8.5, pi 1.8RC7

[pfile]

no details because it was one of those "crash without coredumps"... the only trace of the crash is this message in syslog:


10/6/13 3:13:40.556 PM com.apple.launchd.peruser.501[349]: ([0x0-0x6ad6ad].com.pleiades-astrophoto.PixInsight[6952]) Exited with code: 255


the crash had something to do with going back and forth in the history. i had an image with a big black border, so to look at the histogram properly i made a preview over the center area and told HT to track the preview. of course when undoing the preview went away, so i made another preview (which had the same name). i guess since the two preview objects are different objects despite having the same name, the histogram viewer would lose track of it. i went back and forth like this a few times, recreating the preview, but then i realized i should create the preview 2 steps back...

so i backed up two history steps and made the preview there. when i tried to select the preview in HistogramTransformation, i got the "Critical segmentation error catched" dialog and when i said "OK" to that, PI crashed with no log.

i assume this has to do with the different preview objects getting created and destroyed. i have not yet tried to reproduce it but i will. i still have all the *swp files if those are interesting at all.

rob

[Juan Conejero]

Hi Rob,

I'm sorry to read this. I'm going to investigate this problem, but I need a repeatable pattern to know what happens. This sounds like a structural problem, so it shouldn't be platform-specific. I'll try to reproduce it on all platforms.

Sorry for the trouble.

[pfile]

it's okay. i was able to restore the important images from the .swp files that were hanging around.

i have not had time to reproduce it, but now that my mount is dead i expect to have lots of free time

rob

[pfile]

while trying to reproduce this problem i ran into something that is hopefully related but perhaps not. if you load any image and then just start hitting command-I as quickly as possible, eventually you will get a low-level API abort with invalid module handle.

rob