Bug 16237

Summary:	Should the XCB backend for cairo_surface_flush() call xcb_flush()?
Product:	cairo	Reporter:	junkmailnotread
Component:	xcb backend	Assignee:	Carl Worth <cworth>
Status:	RESOLVED WONTFIX	QA Contact:	cairo-bugs mailing list <cairo-bugs>
Severity:	enhancement
Priority:	medium	CC:	silverpower
Version:	1.6.4
Hardware:	x86 (IA32)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:

Description junkmailnotread 2008-06-05 05:11:21 UTC

Should the XCB backend for cairo_surface_flush() call xcb_flush()?

If I call cairo_paint() on an XCB backend, the source image will not be displayed on the screen until I call xcb_flush().

For reasons of portability - at some point I want to switch to Glitz - I would prefer to call cairo_surface_flush() instead of xcb_flush(). However cairo_surface_flush() has no effect because the XCB backend flush function is NULL.

The description for cairo_surface_flush() says "Do any pending drawing for the surface ...". To my mind that justifies calling xcb_flush(), since otherwise no drawing takes place.

However the description also says "If the surface doesn't support direct access, then this function does nothing." I must admit I'm not sure what that means; perhaps it explains why xcb_flush() is not called.

Comment 1 Behdad Esfahbod 2008-06-05 10:06:05 UTC

No it shouldn't.  We made xlib_flush do that, and it was a disaster.  What the surface_flush() means is that cairo flushes all its internal state out.  So in your case, it's now living in xcb connection cache.

I have proposed in the past to make cairo_surface_show_page() do the xlib/xcb_flush in the past.   Still think that's a useful feature.

Comment 2 Chris Wilson 2008-10-10 11:24:44 UTC

cairo_surface_flush() only guarantees that all operations are complete before you try to read back from the surface. X enforces that all outstanding operations are complete before a XGetImage() reads back from the surface. The remaining issue would then be somebody mixing direct pixel manipulation with XRender to a ShmPixmap and that would require cairo_surface_flush() to call XSync()! And if XFlush() was a performance disaster...

However, cairo_surface_show_page() calling XFlush() has merit and would seem to fit the sematics, so copying to TODO.

Comment 3 Stuart Bennett 2009-01-13 16:22:44 UTC

*** Bug 16031 has been marked as a duplicate of this bug. ***

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.