Further attempts at squashing the catalog scanning crash

[pjrobertson]

I think this has now just become a competition between me and skurfer to see who can create the most commits that are redundant ;-)

I think I'm winning atm :D

---

From #1319:

Nooooooooo!

Looks like this still isn't fixed, 4 crash logs this morning. :(
How is this possible!?

I think perhaps the problem was with the dispatch_async call in saveIndex. I've changed it around slightly now so it only uses dispatch_sync, and if it's being called on the scanQueue queue, then it won't bother trying to add it to the scanQueue again. (If I'm correct, dispatch_async will make a 'copy' of the code/iVars etc. when it's called and add them to memory for when the block can be run, whereas dispatch_sync will wait until the block can be run then make a copy of the code/iVars, so potentially the iVars in the two blocks could be different.

I've also added __block to the indexDate iVar. I remember @skurfer mentioned something about not having to do this if you're using setters, but I couldn't remember the reason why, and I think it's better to be safe

This is a fun exercise right? :D

[pjrobertson]

Hang on hang on...

All the crash logs in the '3940' folder on qsapp.com are version 3939 (it's version 3940 that uploaded the crash reports), so we're OK :)

Still, I think this should still be merged. Calling saveIndex on the scanQueue blindly from the scanQueue (called in scanAndCache might perhaps cause a lock)

[skurfer]

Still, I think this should still be merged.

Agreed.

Should we go ahead and change line 529 to runOnQueueSync as well?

[skurfer]

Oh, and why don't you add the latest from Transifex as long as we're merging another pull. :-)

[pjrobertson]

OK, change pushed.

Turns out there were no Transifex files to pull (no new files over 80%
complete) so that's all fine :)

On 15 January 2013 13:42, Rob McBroom notifications@github.com wrote:

Oh, and why don't you add the latest from Transifex as long as we're
merging another pull. :-)

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/1321#issuecomment-12267061.

[skurfer]

Hopefully most people are on weekly updates and never even saw 3940 (or 3939). :-)