[RFC Weston 00/10] Sub-surfaces v2

[Pekka Paalanen]

Hi all,

this is a new spin of the sub-surfaces Wayland protocol
extension, v2. The v1 was announced in:
http://lists.freedesktop.org/archives/wayland-devel/2012-December/006844.html

Earlier we have already landed some 14 patches of toytoolkit
restructuring, that was done after v1.

The major changes in v2 are:

- Moved protocol from Wayland to Weston. It is easier to handle
  in one place, and will be moved into Wayland core once it is
  deemed good enough.

- Protocol documentation is updated. The following requests have
  been added to the protocol: wl_subcompositor.destroy,
  wl_subsurface.destroy, and most importantly
  wl_subsurface.set_commit_mode.

- Clients can now choose between two commit behaviours for
  sub-surfaces: parent-cached and independent. These are
  explained in the protocol file.

- A totally new demo client presenting sub-surfaces, including
  Cairo-image rendered window with a pure EGL/GL widget in a
  sub-surface, running independently, but still without glitches
  on resize (sans bugs).
  http://people.collabora.com/~pq/subsurface-2.png

The biggest improvement over v1 is that we have some thought-out
commit behaviours. It is possible to resize a window so that all
surfaces stay in sync on screen, and it is also possible to have
sub-surfaces running on their own (i.e. without commit the
parent surface, too).


Open issues:

- Does this commit behaviour satisfy all needs?

- Input events still very much unexplored. The demo just punts
  them.

- Clipping and scaling of surface content - this will most
  likely be a new orthogonal extension altogether, so not really
  relevant here.

- What should wl_subsurface.destroy do? Nothing, or reset the
  wl_surface's role? Currently it resets the role, since it was
  easy to implement.

- Should sub-surface nesting be allowed? I.e. creation of
  sub-sub-surfaces etc. What would be semantics be? I have not
  really thought about that. Input dispatching within clients
  may be a problem.

- wl_shell and other protocols interact with sub-surfaces, and
  all interactions have not been resolved yet.

If I forgot any other design issues, let me know, please.


To do, bugs:

- destroying the wl_surface of a wl_subsurface will prevent
  destroying the wl_subsurface, fix this, and adjust the
  protocol description accordingly

- double-buffering of sub-surface z-order changes

- a demo client with window decorations stitched from 4
  non-overlapping sub-surfaces

- fix frame throttling in subsurfaces demo

- fix eglSwapInterval in Mesa (krh?)

- subsurfaces demo behaves strangely under the DRM backend of
  Weston, debug and fix it. A timestamp issue?

- even with three buffers in a pool, subsurfaces may still run
  out of free buffers. When that happens, "all buffers are held
  by the server" error message appears, and sub-surfaces may be
  out of sync. Related to frame throttling.

- glitched rendering of the GL widget in subsurfaces when Up or
  Down key held pressed, might be related to frame throttling.

- fix full-surface alpha for surfaces that have sub-surfaces

- in subsurfaces demo, put the GL widget into its own thread, so
  it is truely as concurrent as it can get.

- reimplement the "black surface" in shell as a sub-surface


The new code can be browsed in:
http://cgit.collabora.com/git/user/pq/weston.git/log/?h=subsurface-v2

Pekka Paalanen (10):
  protocol: add sub-surfaces
  compositor: introduce sub-surfaces
  tests: add sub-surface protocol tests
  compositor: add sub-surfaces to repaint list
  shell: keyboard focus and restacking fixes for sub-surfaces
  window: implement shm triple-buffering
  window: create sub-surfaces
  window: implement per-surface redraws
  clients: add subsurfaces demo
  window: prevent EGL sub-surface deadlock

 clients/.gitignore      |   3 +
 clients/Makefile.am     |  12 +
 clients/subsurfaces.c   | 791 ++++++++++++++++++++++++++++++++++++++++++++++++
 clients/window.c        | 343 ++++++++++++++++++---
 clients/window.h        |  14 +
 configure.ac            |   3 +
 protocol/subsurface.xml | 214 +++++++++++++
 src/.gitignore          |   3 +
 src/Makefile.am         |   4 +
 src/compositor.c        | 689 ++++++++++++++++++++++++++++++++++++++++-
 src/compositor.h        |  59 ++++
 src/shell.c             |  37 ++-
 tests/.gitignore        |   3 +
 tests/Makefile.am       |   9 +-
 tests/subsurface-test.c | 327 ++++++++++++++++++++
 15 files changed, 2437 insertions(+), 74 deletions(-)
 create mode 100644 clients/subsurfaces.c
 create mode 100644 protocol/subsurface.xml
 create mode 100644 tests/subsurface-test.c


This patch series depends on further patches, so you cannot just
apply this series on top of Weston master. I did not include the
other patches, because they are waiting for a rewrite. Instead,
you can use my Weston branch subsurface-v2 which has everything
in place. You do not need any patches for Wayland, you can just
use master of Wayland upstream.

The following changes since commit a4575634f4a1f63430cfc4da5b7ab757f6c33c1c:

  compositor: Rename lock and unlock signals to idle and wake (2013-02-21 21:12:45 -0500)

are available in the git repository at:

  git://git.collabora.co.uk/git/user/pq/weston.git subsurface-v2

for you to fetch changes up to 2b3f13a2cd8ddadd87b5329e400fcddcb460a046:

  window: prevent EGL sub-surface deadlock (2013-02-22 15:47:03 +0200)

When you build it, you should configure Weston with
--with-cairo-glesv2 regardless of whether your Cairo actually
has GLESv2. Otherwise the new subsurfaces demo might not be
built.


Thanks,
pq

-- 
1.7.12.4