Bibledit Lite (was: Re: [be] Packages (bibledit-web and build time mysql server checking))

[Dan Dennison]

On Tue, Dec 28, 2010 at 5:36 PM, Jonathan Marsden <address@hidden> wrote:

I started on this today.  I soon encountered something I am not sure how
to handle -- bibledit-web seems to expect a working and configured mysql
server at *build* time.

I actually second this view. The buildbot at buildbot.biblesint.org used to work from Bibledit HEAD, but stopped working once a running mysqld was required at build time. We decided it would be easier to fork into Bibledit Lite than to try and track a bunch of changes for what is now server-code, when all we want is the GTK+ client.

That said, we do have Bibledit Lite for Windows downloadable at http://buildbot.biblesint.org. It's not quite where we want it, notably it still needs a better answer for USB repositories, but it's working a lot better than Bibledit for Windows ever has, and a number of heap corruptions and other more-Windows specific issues have been address. Bibledit Lite is a fork of Bibledit from just before the detached windows patch went in, so it has the old paned/tabbed UI. Bibledit Lite also uses Git instead of snapshots.

Teus, if you want to patch in my changes for config.xml.cpp feel free. There were several failure-to-xmlFree cases in there. I've noted these problems are still in 4.2. Once those were fixed, the code seemed to use a lot less RAM. We've seen running Bibledits under 25MB RAM. Bibledit on Windows used to use > 50MB RAM. We also fixed the keyterms import causes crash, which happened because the keytermsgui object wasn't initialized. There were a bunch of other crash fixes as well.

Our next area of code to fix is screen.cpp, to solve the random crashes that happen when the GTK+ text buffer changes while we're trying to scroll to a reference, as well as a few edge cases such as what happens when you scroll to verse 0 or chapter 0.

Looking ahead, we plan to kill the verse tracker in its polling form and move it to event-driven. This should reduce the background CPU usage of Bibledit to almost zero. Further out is checkpointing of Git state on every change to the queue, rather than at exit. This way a crash will not cause the pending Git tasks to be lost, rather to be pending it must write out gitjobs.xml.

The big mysteries we're trying to solve right now is the screen jumping to the left on restore/maximize, and various screen corruption bugs that happen when window elements go out of bounds.

dan :)