Re: [Gnash-dev] Releasing 0.7.2 RSN (i.e. on monday)...

[Rob Savoye]

strk wrote:


> I've tried to higher severity of bugs that should be show stoppers
> for 0.7.2 and I ask you to do the same. Brows the current bugs
> database and set to "Blocker" any item that you think should block
> 0.7.2.

  At this point I'd say a blocker is anything that causes a core dump.

> Currently we have two: #18207 (the sound thing) and #18100 (stack
> corruption). 

  #18207 I think we should fix, as it turns out hitting Escape for any
movie causes Gnash to core dump. The problem is in gtk.cpp where it
handles control keys. I just checked in a workaround so at least it
doesn't core dump. Esc was being handled like a control key, and it
turns out that calling gtk_main_quit() from a key handler instead of a
menu item is a bad thing. For now control keys are handled better, but
"quit" from the keyboard doesn't work. I assume most people would use
the menu item to quit, which works just fine. The keys were mostly only
used for SDL, since it has no menu.

  #18100 is much harder to track down, since most of us aren't working
on OpenBSD machines. We may have to make a release note about stack
protection, and then try to fix it in CVS.

> So, let's define what a show-stopper is. Bear in mind that we're
> threaded now, so memory corruptions are likely waiting around any
> corner...

  At this point a code freeze would mean no additional changes that
aren't required. I'm guilty of sliding things in too, like the cross
configuration changes. At this point we should concentrate on core dump
bugs, or we'll never get this out.

  One thing to be careful of is making sure new/delete or malloc/free
are thread save by always wrapping memory allocation in a thread with a
mutex.

        - rob -