GNOME Bugzilla – Bug 787900
Don't allocate actors a negative size
Last modified: 2017-09-25 16:47:33 UTC
Clutter will loudly complain when this happens. By booting up without a display, stage dimensions are typically too small, and it exacerbates some underlying minor bugs in the allocation code. From a patch we currently carry around on Endless downstream: https://github.com/endlessm/gnome-shell/commit/d07ff590703eb148897f13e3276d137794c397c4
Created attachment 360050 [details] [review] Don't allocate actors a negative size Attaching patch (by Cosimo Cecchi)
Created attachment 360052 [details] [review] Don't allocate actors a negative size
(In reply to Mario Sánchez Prada from comment #0) > By booting up without a display, stage dimensions are typically too > small, and it exacerbates some underlying minor bugs in the > allocation code. Does this still apply with the patches from bug , in particular https://git.gnome.org/browse/mutter/commit/?id=0aa7405a2a895719fb7365d?
Review of attachment 360052 [details] [review]: I'm not sure that not handling an allocation that is smaller than the requested minimum size is a bug. But if this is something we do want to take into account, then I doubt those two places are the only ones affected - for a starter, your patch only touches the LTR code paths ...
(In reply to Florian Müllner from comment #3) > (In reply to Mario Sánchez Prada from comment #0) > > By booting up without a display, stage dimensions are typically too > > small, and it exacerbates some underlying minor bugs in the > > allocation code. > > Does this still apply with the patches from bug , in particular > https://git.gnome.org/browse/mutter/commit/?id=0aa7405a2a895719fb7365d? That's interesting, it could be very well not a problem anymore right. I'm not at home right now, but I will test it later today without this patch and report here. (In reply to Florian Müllner from comment #4) > Review of attachment 360052 [details] [review] [review]: > > I'm not sure that not handling an allocation that is smaller than the > requested minimum size is a bug. But if this is something we do want to take > into account, then I doubt those two places are the only ones affected - for > a starter, your patch only touches the LTR code paths ... I'm not sure either, let me check this in the actual hw first and I'll let you know. This patch is pretty old now, and it might not be needed anymore after all.
Tested in a actual machine but I can't tell whether the patch makes a difference or not because it's already crashing on mutter when booting without any display attached. Need to investigate that first, it seems..
It looks like what I needed was a bunch of patches from bug 730551 (we don't use Wayland for now in Endless, so I could skip those) + patch the shell accordingly not to assume a monitor, as it does right now. However, such a patch for the shell is basically the one from Jonas available in 730551 which, for some reason, hasn't landed yet (overlooked maybe?), so I'm closing this bug as a duplicate of that one. *** This bug has been marked as a duplicate of bug 730551 ***