1641778 - [Wayland] moz_container_wayland_get_surface needs to be thread-safe

Status: RESOLVED FIXED

(Core :: Widget: Gtk, defect)

Description

[Martin Stránský [:stransky] (ni? me)]

moz_container_wayland_get_surface needs to be thread-safe and we should lock/unlock the acquired surface.

Comment 1

[Martin Stránský [:stransky] (ni? me)]

Or better we should stop export moz_container_wayland_get_surface() but create an api for particular code like set frame callback etc.

Comment 2

[Kenny Levinsen :kennylevinsen]

What's the reason for us needing to wait for a frame callback on the parent to create a surface in the first place?

Splitting get_surface into create and get instead of just making it thread-safe would probably be cleaner than the slightly odd auto-creation thing we have going on right now. If we could get creation done without waiting for a frame, we'd be able to get things on screen a frame faster, and we would probably be able to kill the "initial_draw_cb" thing.

Comment 3

[Martin Stránský [:stransky] (ni? me)]

(In reply to Kenny Levinsen :kennylevinsen from comment #2)

What's the reason for us needing to wait for a frame callback on the parent to create a surface in the first place?

Splitting get_surface into create and get instead of just making it thread-safe would probably be cleaner than the slightly odd auto-creation thing we have going on right now. If we could get creation done without waiting for a frame, we'd be able to get things on screen a frame faster, and we would probably be able to kill the "initial_draw_cb" thing.

Because we create a subsurface of the parent surface and we can't create it before the parent is also created.

Comment 4

[Kenny Levinsen :kennylevinsen]

It should only exist, though. We shouldn't need to wait a frame - the spec doesn't even seem to require the parent to be configured. I should look into when Gtk actually creates the surface...

It's a bit annoying to have to delay execution of these things, but reducing get to a simple accessor and keeping all the creation logic separate would probably still be nice.

Comment 5

[Martin Stránský [:stransky] (ni? me)]

Sure, we can reiterate that again and try to make it work.

Comment 6

[Martin Stránský [:stransky] (ni? me)]

Attachment: Bug 1641778 [Wayland] Always lock wayland surface when se use it, r?jhorak

• moz_container_wayland_get_surface() is replaced by moz_container_wayland_surface_lock()/moz_container_wayland_surface_unlock()
  pair which ensures the surface is locked and won't be destroyed by other thread when window is hidden for instance.

• nsWindow::GetWaylandSurface() is removed and nsWindow::GetMozContainer() needs to be used instead.

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/9b17e776c117
[Wayland] Always lock wayland surface when se use it, r=jhorak

Comment 8

[Cristian Brindusan [:cbrindusan]]

https://hg.mozilla.org/mozilla-central/rev/9b17e776c117