Subsurface-mobile support request

[Dirk Hohndel]

On Fri, 2020-04-03 at 15:01 -0700, Linus Torvalds wrote:
> On Fri, Apr 3, 2020 at 2:37 PM Dirk Hohndel via subsurface
> <subsurface at subsurface-divelog.org> wrote:
> > I need to spend more time staring at the code to try to figure out
> > how
> > it's possible that we find two new dives but then believe that the
> > dive
> > list is unchanged - as THAT is the underlying problem here. Which
> > is
> > different from the problem we had with 3.0.x with x<3 where we
> > thought
> > that there was already a write ongoing and refused to start a
> > "second"
> > write...
> 
> Hmm. "move_dive_table()" (which does the "move from download thread
> to
> main divelist) doesn't actually seem to do mark_divelist_changed().

So the UI code uses the undo mechanism to do this and it seems that
either part of this code assumes that the other part of it marks the
divelist changed. I'm still trying to make sure that I don't break
unrelated things here, but I believe that I have identified where this
goes sideways. The question now is where exactly to fix it.

> So maybe that never happens, and then unsaved_changes() returns
> false,
> and then we don't even try to save anything?

Yep.

> Back to merging kernel ARM code for me..

You get to have all the fun :-)

/D