Comment 10 for bug 804936

Darin,

I would look at zim/fs.py and than class WindowsFile, methods _on_write and _recover should be first point of attention. Just dump all state there. If we can reconstruct the state of the .new and .bak files when this happens it might become a bit clearer.

Also you might want to put some output on start and end of write_async and writelines_async in the UnixFile class. Although the lock should prevent two async operations happen at the same time, it is worth checking.

Btw. missing detailed error log of what error happens in fs.py. Probably goes to error log in %TMP%/zim.log . If not you might want to check the code that generates the exception and have it print the original exception.

Regards,

Jaap