Qt branch updates

Dirk Hohndel dirk at hohndel.org
Sun May 12 07:12:40 PDT 2013


On Sun, 2013-05-12 at 06:31 -0700, Dirk Hohndel wrote:
> On Sun, 2013-05-12 at 10:18 +0200, Henrik Brautaset Aronsen wrote:
> > Dirk Hohndel wrote:
> > > long time Subsurface users... please try this in the Qt branch and make
> > > sure I got all the corner cases right (I'd be very surprised if I did,
> > > to be honest - but I tried reasonably hard).
> > 
> > Crash:
> > 
> > 1) ./subsurface dives/*xml
> > 2) Select the first dive
> > 3) Arrow down
> > 4) BOOM:
> > 
> > Thread 0 Crashed:: Dispatch queue: com.apple.main-thread
> > 0   QtGui                             0x0000000102318a58 
> > QGraphicsItem::pos() const + 8
> > 1   subsurface                        0x0000000101c721da 
> > QGraphicsItem::y() const + 24 (qgraphicsitem.h:282)
> > 2   subsurface                        0x0000000101c66f91 
> > ProfileGraphicsView::plot(dive*) + 3217 (profilegraphics.cpp:319)

> Ha. I clearly wasn't as awake last night as I thought. I apparently
> didn't push my changes... this makes no sense... I need to check this
> one more time (this would allow me to improve the commit message on the
> commit that I thought I had already pushed... and which I checked that I
> had pushed and that it was too late to change...)
> 
> Must be the drugs and the alcohol.

Or just being tired and frustrated after wasting all the time on the
stupid Gtk/libsoup/glib-networking/Windows crap...

Anyway. I was on a different branch and did a git push Qt. Then I saw
that I had forgotten to split the selection commit and do a good commit
message. And decided that I was done for the night and sent the email
that you responded to.

But of course having been on a different branch, the git push Qt did
nothing... Duh.

So I now split the commit in two and created useful commit messages. And
then applied those two commits plus two bug fixes to the Qt branch and
pushed this all out.

And one of the two fixes addresses the crash that you ran into above...
that crash was Tomaz' doing. An object was only initialized in some
cases (when we had a valid nickname or model name for a dive computer);
and then got unconditionally dereferenced in the very next line of code.

All better now.

Which brings me back to my earlier two requests - now with actual code
to test in the git tree:

Subsurface users - test that the selection seems right.

Qt people - tell me all the ways in which my selection code is an
abomination and broken.

Thanks

/D



More information about the subsurface mailing list