kicad-developers team mailing list archive

[Chris Pavlina <pavlina.chris@xxxxxxxxx>]

On Thu, Mar 23, 2017 at 08:25:27PM -0400, Jon Evans wrote:
> That sounds like a reasonable approach if the lag time of running the test
> is low enough to not be a nuisance.  Are these crashes you speak of ones
> that appear when the user first tries to switch to OpenGL, or later?

Ideally, I'd like kicad.exe to start it in the background immediately,
so by the time you load a project and click over the pcbnew it has
already run. Of course if you start pcbnew in standalone it can't help
that, but I think most users generally don't.

Crashes are generally immediate, either OpenGL works properly or it
doesn't.

> 
> One alternative approach (that might actually work well alongside yours) is
> a crash sentinel.
> Basically if you save some value somewhere that indicates a clean exit of
> the program, then if you start the program and the value isn't as expected,
> you can assume that the program crashed, and take some action (for example,
> falling back to Cairo canvas as a "safe mode" and alerting the user)

I've wanted some form of crash detection (and reporting?) for a while
but it's way down my priority list to implement. I'm much more concerned
with preventing them from happening. Remember, once a crash has actually
occurred in the main program you're already past the point of no return
for possible data loss.

> 
> -Jon
> 
> On Thu, Mar 23, 2017 at 8:11 PM, Chris Pavlina <pavlina.chris@xxxxxxxxx>
> wrote:
> 
> > Hi,
> >
> > I've been thinking about my intent to make KiCad fall back cleanly to
> > Cairo when OpenGL isn't available. From what I can see, there are
> > certain situations where trying to use OpenGL causes crashes that appear
> > very hard to prevent.
> >
> > I really want to do something about this - they're hard to prevent, but
> > I still don't think it's acceptable to crash out on the user. What if we
> > compiled a small test binary that just tried to load an OpenGL canvas?
> > We should be able to test for functionality pretty easily by starting
> > that executable, and any crash would be confined to its own process. Is
> > anyone opposed to this? I know it's not the cleanest solution, but I
> > can't think of much better.
> >
> > --
> > Chris
> >
> > _______________________________________________
> > Mailing list: https://launchpad.net/~kicad-developers
> > Post to     : kicad-developers@xxxxxxxxxxxxxxxxxxx
> > Unsubscribe : https://launchpad.net/~kicad-developers
> > More help   : https://help.launchpad.net/ListHelp
> >