Bug #923

Tony build 0.5 crashes on re-load of .ton file

Added by Matthias Mauch over 8 years ago. Updated over 8 years ago.

Status:NewStart date:2014-04-03
Priority:HighDue date:
Assignee:-% Done:


Target version:-


Hm, I think I got a stack trace... but now it just crashes without
any feedback.

I've removed the 0.5 downloads for now, since this wasn't really
going to work. We do have the 0.4 downloads, so that's cool,
I think.

AClassicEducation_NightOwl_premix_08.ton - corresponding audio 72.5 KB, downloaded 44 times Rachel Bittner, 2014-04-09 05:50 PM

AClassicEducation_NightOwl_premix_08.wav 28.8 MB, downloaded 7 times Rachel Bittner, 2014-04-09 05:50 PM


#1 Updated by Chris Cannam over 8 years ago

Fixed in 8eb167e817e7

Sessions saved from prior versions may still not reload correctly if they had extant selections (with visible pitch candidate tracks) at the time you saved them. Sessions saved from this version should work correctly.

(The crash should be fixed in any case, but there was another problem with the saved sessions as well: they should not have included the candidate tracks)

#2 Updated by Rachel Bittner over 8 years ago

Chris- could I trouble you to make a v0.6 build that includes this fix? The annotators have made this mistake a few times.

#3 Updated by Chris Cannam over 8 years ago

I actually replaced the v0.5 build with another one that has this fix in it. That is, the current v0.5 on the downloads page does have this fix.

(I know it's very bad form to replace a binary when it's already up for download, but in this case I was pretty confident that nobody but Matthias and I had downloaded it before he pulled it and I replaced it. I should have remembered the paper trail in issues like this one though.)

So the v0.5 now available should not crash on reloading a .ton session. If it does, then it's a bug I don't know about -- if you can attach an example session that causes the crash, that would be great (my time is very limited and I may not be able to fix it quickly enough for you but it would be a good piece of information to have anyway).

[Edit: if the problem is that the v0.5 release doesn't properly reload sessions saved from v0.4, then I'm afraid that's a different problem and it's a bug in v0.4 -- it isn't saving all the information necessary to reload the session. So it can't actually be fixed retrospectively. If you have sessions that are affected this way, it is possible to edit the session files by hand if desperate]

#4 Updated by Rachel Bittner over 8 years ago

No sessions from 0.4, but I do have sessions saved with the bug from the old v0.5. Is there any way to salvage them? If need be could send you the "broken" .ton files.

#5 Updated by Rachel Bittner over 8 years ago

Chris - a quick update to let you know that we've managed to salvage the files and all is well.

#6 Updated by Rachel Bittner over 8 years ago

A new, possibly worse bug has just appeared. Files saved with pitch candidates visible can now be loaded, but no changes can be made and any exported .csv files are blank. Example .ton file + audio attached.

#7 Updated by Matthias Mauch over 8 years ago

Hi Rachel. Ok, this clearly is a bug, but I've played around a bit myself, and I don't encounter the problem when I save "normally", i.e. with no pitch candidates showing. Is this your impression as well? If so, I think you can probably just instruct your annotators not to save with pitch tracks showing.

The reason I'm saying this is that Chris told me he was giong to be on holidays for the next two weeks from today, so I don't think we can do much on this problem till he's back. I will try to tackle some of the other issues though.

#8 Updated by Rachel Bittner over 8 years ago

Hi Matthias,

Yes - this isn't a problem if you save "normally", and this is not a problem moving forward.

My issue right now is that I have annotation files that were completed by my annotators, but were saved in the "bad" way, and we can't get the output from those annotations (an example being the one I attached earlier). Exporting to csv returns blank or almost blank files (only will output the "yellow" pitch track), and doing a save-as doesn't fix the issue.

I think the problem is in how tony loads those bad sessions. I think there is probably a way to hack it so that the problem files load correctly. I looked at the source code and I think the issue is in svapp/framework/SVFileReader.cpp
in particular where the layers get loaded in (lines 787-910). I tried poking around with it a little but I'm not that familiar with the SV framework.

Quick note to say that the other issues I filed last night are not at all urgent, but were more to make note of for future versions.

#9 Updated by Rachel Bittner over 8 years ago

  • Priority changed from Urgent to High

Also available in: Atom PDF