Impress crashes frequently on many different files, either during editing or just randomly.
I have re-initialized my user profile, it might have helped for a few hours but then the problem started again.
Please advise how I can provide more useful info, such as a crash dump.
Platform: Ubuntu 10.10.
tag libreoffice-126.96.36.199, Ubuntu package 1:3.3.0-1maverick1
It would be helpful to attach your backtrace log when meeting a crash, please refer to:
I raised up the importance of this bug because I can nearly ALWAYS reproduce it in 10 minutes' using of Impress. But I could not get a exact steps of reproducing. What all I did to make the issue appearing are mostly switching my behavior in these three areas in an newly created empty presentation:
a. Switch to Normal tab, Add title in a slide, Add body text in a slide
b. Switch to Notes tab, Add some Notes
c. C-p to call out the printing dialog then cancel it
The crash will happen unpredictably(so far..) in any of the area.
I am attaching a backtrace when reproducing this.
Created attachment 44467 [details]
Well, referring to the bug:
Find the always reproducible steps by chance:
1. Launch Impress and create an empty presentation
2. C-p to open the print dialog
3. Close the dialog by clicking 'Cancel'
4. Switch to Outline tab
5. entering a character, "1" for example
=> Libreoffice Crash
Created attachment 44475 [details]
one plausible fix
So, when you activate the print dialog we get a print renderer which wants a DrawView. The print renderer stays around for the life of the ViewShellBase. The DrawView takes the current DrawViewShell, but that's destroyed when flipping when flipping views.
So, we either use a DrawView with a NULL DrawViewShell, or we take another reference to the current DrawViewShell and ensure it exists for the lifecycle of the DrawView.
It doesn't *seem* to be the case that we actually need a DrawView with an attached DrawViewShell.
Created attachment 44557 [details]
Yeah, patch looks right for -3-3 surely, slightly amended to have the last unchecked use of that ptr covered.
@caolan, if you could ack that, I'll shove it into -3-3
looks good +1
as an aside I can see from our own auto-crash reports that we (and OOo) appear to have a similar problem in writer somewhere
*** Bug 35273 has been marked as a duplicate of this bug. ***
this isn't in master, and it isn't in 3-4
Meh. Thought there was a ~regular merging going on ...
yeah, pmladek said he'd sync up unmerged 3-3 changes to master and 3-4 today which should recover this