Bug 161904 – [GTK] Drop redundant wl_display_flush_clients() call

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKitGTK

[Emanuele Aina]

The Wayland GSource::prepare() function already calls wl_display_flush_clients() at every mainloop iteration, so there's no need to further call it on Surface::commit().

[Emanuele Aina]

Created attachment 288684 [details]
Patch

[Carlos Garcia Campos]

Comment on attachment 288684 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=288684&action=review

> Source/WebKit2/ChangeLog:10
> +        The Wayland GSource::prepare() function already calls
> +        wl_display_flush_clients() at every mainloop iteration, so there's no
> +        need to further call it on Surface::commit().

That's not equivalent, prepare is called before dispatch, and commit happens on dispatch. I'm not saying we need to flush on commit, because I don't know, I copied that from previous patches, but this explanation is not clear to me.

[Emanuele Aina]

Yep, I didn't mean to imply that it was equivalent, just that as far as I can tell, even by asking to Wayland developer around here, there's no reason to call wl_display_flush_clients() in ::commit(), and that the call in ::prepare() is the place where they expected the call to be.

For further reference, the Wakefield experimental compositor only does it in the ::prepare() function: https://github.com/alexlarsson/wakefield/blob/master/wakefield-compositor.c#L1919

Let me know how I can make the ChangeLog entry more clear.

[WebKit Commit Bot]

Comment on attachment 288684 [details]
Patch

Clearing flags on attachment: 288684

Committed r206961: <http://trac.webkit.org/changeset/206961>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.