AUDIT: r21825 - gnucash/trunk/src/backend/sql - [Bug 666329] Empty database after a little while
jralls at ceridwen.us
Sun Jan 8 23:43:34 EST 2012
On Jan 8, 2012, at 3:21 PM, John Ralls wrote:
> On Jan 8, 2012, at 1:20 PM, Christian Stimming wrote:
>> Great!!!!!!! Thanks, John, for tracking down this mysterious and worrysome bug
>> for all of our SQL users. This is very very good work!
>> I've copied the commit to 2.4 as well. As this was the last outstanding issue
>> before 2.4.9, we can tag the branch with 2.4.9 tonight, right?
>> Thanks a lot!
> Oh, I'm not done yet... and I've no confidence that that's the only path that could allow a timed autosave to sneak through. I've got part 2 done (changing from gnc_dbi_sync_all to gnc_dbi_safe_sync_all), though it's not in subversion yet, and I'm working now on separating the book_dirty (which applies to the book object and is needed to get it saved when one of its properties or KVPs gets edited) and session_dirty (which applies to everything). The Xml backend's commit_edit method can call the latter, and that can fire off the autosave function (I wanted to do everything in the Xml backend, but the autosave function has too many hooks into gnome and qof, so I decided to leave it where it is for now and just eliminate the overloading of book_dirty.)
> If it goes well, I should have it checked in, backported and a make distcheck running, and we can bump and tag the release tomorrow.
But unfortunately it didn't go well. The autosave fires once, but subsequent edits aren't causing the FileSaveAction to be enabled (plus it's enabled at startup, but that was going on before I started). More work tomorrow.
More information about the gnucash-devel