Merge lp:~andreas-pokorny/mir/nbs-resize-first-stream-of-surface into lp:mir

[Andreas Pokorny]

Make NBS behave like OBS - resize the first buffer stream

With OBS mir::scene::BasicSurface would pull out the first stream as the 'default' stream and forward all resizes to that stream. With NBS this is done by the client library. So if no buffer stream is configured via the mir_surface_spec_add_buffer_stream or mir_surface_spec_set_streams API the client side surface will forward the content of resize events to the 'default' stream.

Since we deprecate using mir_surface_get_buffer_stream users are forces to specify the streams or presentation queues they want for display. But streams buffers lack a resize functionality so switching to the new function would not be a viable option.

With presentation chains the client handles resizes by submitting buffers in an adjusted size. With BufferStream there is no way a client can resize an existing stream. With OBS the server made sure that upcoming allocations adapt to the window size. Since the BufferStreams functionality moved to mirclient with NBS, the resize is forwarded to the 'default' BufferStream.

This change makes sure that even when the caller uses mir_surface_spec_set_streams or mir_surface_spec_add_buffer_stream - the 'default' or in other words first stream still gets resized.

[lp:~andreas-pokorny/mir/nbs-resize-first-stream-of-surface]

3616. By Andreas Pokorny on 2016-07-27

spell out the actual expectation: no exception when attempting to resize a presentation chain

[Mir CI Bot (mir-ci-bot)]

FAILED: Continuous integration, rev:3615
https://mir-jenkins.ubuntu.com/job/mir-ci/1357/
Executed test runs:
    FAILURE: https://mir-jenkins.ubuntu.com/job/build-mir/1617/console
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-0-fetch/1670
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=vivid+overlay/1661
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=xenial+overlay/1661
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=yakkety/1661
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1632
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1632/artifact/output/*zip*/output.zip
    FAILURE: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial+overlay/1632/console
    FAILURE: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1632/console
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1632/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1632
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1632/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial+overlay/1632
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial+overlay/1632/artifact/output/*zip*/output.zip

Click here to trigger a rebuild:
https://mir-jenkins.ubuntu.com/job/mir-ci/1357/rebuild

review: Needs Fixing (continuous-integration)

[Mir CI Bot (mir-ci-bot)]

FAILED: Continuous integration, rev:3616
https://mir-jenkins.ubuntu.com/job/mir-ci/1358/
Executed test runs:
    FAILURE: https://mir-jenkins.ubuntu.com/job/build-mir/1618/console
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-0-fetch/1671
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=vivid+overlay/1662
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=xenial+overlay/1662
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=yakkety/1662
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1633
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1633/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial+overlay/1633
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial+overlay/1633/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1633
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1633/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1633
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1633/artifact/output/*zip*/output.zip
    FAILURE: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial+overlay/1633/console

Click here to trigger a rebuild:
https://mir-jenkins.ubuntu.com/job/mir-ci/1358/rebuild

review: Needs Fixing (continuous-integration)

[Mir CI Bot (mir-ci-bot)]

PASSED: Continuous integration, rev:3616
https://mir-jenkins.ubuntu.com/job/mir-ci/1359/
Executed test runs:
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-mir/1621
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-0-fetch/1674
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=vivid+overlay/1665
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=xenial+overlay/1665
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=yakkety/1665
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1636
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1636/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial+overlay/1636
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial+overlay/1636/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1636
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1636/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1636
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1636/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial+overlay/1636
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial+overlay/1636/artifact/output/*zip*/output.zip

Click here to trigger a rebuild:
https://mir-jenkins.ubuntu.com/job/mir-ci/1359/rebuild

review: Approve (continuous-integration)

[Alan Griffiths (alan-griffiths)]

I'm not sure enough of the implications to know if this is a good idea. What user visible feature does it effect?

Also "with_potential_stream_do" seems verbose, maybe "if_stream_do"?

review: Abstain

[Andreas Pokorny (andreas-pokorny)]

> I'm not sure enough of the implications to know if this is a good idea. What
> user visible feature does it effect?
>
> Also "with_potential_stream_do" seems verbose, maybe "if_stream_do"?

Created a bug (lp:1608457) for the problem I observerd with gtk+

[lp:~andreas-pokorny/mir/nbs-resize-first-stream-of-surface]

3617. By Andreas Pokorny on 2016-08-01

rename new method

[Kevin DuBois (kdub)]

I think this is fine as a workaround while the definitions are improved in this area.

The physical buffer size is something the client sets, and the onscreen size is something server/shell sets.

The legacy problem has to do with how the sizes (physical and on-screen) were assumed to be the same. The default stream used to just be resized, and now the 1st stream will just be resized.

A function like:
mir_buffer_stream_set_buffers_size() //sets just the physical buffer size.
might be what's needed to make things in the non-legacy world a bit nicer and clearer.

[Mir CI Bot (mir-ci-bot)]

PASSED: Continuous integration, rev:3617
https://mir-jenkins.ubuntu.com/job/mir-ci/1380/
Executed test runs:
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-mir/1661
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-0-fetch/1714
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=vivid+overlay/1705
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=xenial+overlay/1705
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-1-sourcepkg/release=yakkety/1705
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1676
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=clang,platform=mesa,release=vivid+overlay/1676/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial+overlay/1676
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=amd64,compiler=gcc,platform=mesa,release=xenial+overlay/1676/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1676
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=cross-armhf,compiler=gcc,platform=android,release=vivid+overlay/1676/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1676
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=android,release=vivid+overlay/1676/artifact/output/*zip*/output.zip
    SUCCESS: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial+overlay/1676
        deb: https://mir-jenkins.ubuntu.com/job/build-2-binpkg-mir/arch=i386,compiler=gcc,platform=mesa,release=xenial+overlay/1676/artifact/output/*zip*/output.zip

Click here to trigger a rebuild:
https://mir-jenkins.ubuntu.com/job/mir-ci/1380/rebuild

review: Approve (continuous-integration)

[Alan Griffiths (alan-griffiths)]

*Needs Discussion*

It seems that the client API lacks a mir_buffer_stream_set_size() and adding that would be a better fix than this proposal.

But I'm concerned that clients might want to make an atomic update of size and scale. Is the "correct" approach to create a new buffer stream?

review: Needs Information

[Andreas Pokorny (andreas-pokorny)]

> *Needs Discussion*
>
> It seems that the client API lacks a mir_buffer_stream_set_size() and adding
> that would be a better fix than this proposal.
>
> But I'm concerned that clients might want to make an atomic update of size and
> scale. Is the "correct" approach to create a new buffer stream?

Existing behavior with and without mir_surface_spec_set_streams was that the server controls the size of the first buffer stream and does not care about the other. At the moment the server decides based on the --nbuffers=3 vs --nbuffers=0 whether the first buffer stream is resized or not.

The release-create-with-new-size approach led to hang ups and allocation failures - and those are fixed here:
lp:~andreas-pokorny/mir/fix-buffer-release-requests-after-buffer-stream-destruction

The recreate approach has the advantage that it is immediately clear that the first buffer you get from the buffer stream will have the right size. Additionally the screen update would be tied to the necessary mir_surface_apply_spec call. While with the _set_size approach there are arguments to have it happen instantly or only after the next buffer swap..

But a current implementation detail makes the recreate approach a bit more resource intense since on creation three buffers are allocated. With a sequence of resize events two thirds of all created buffers are just allocated and freed without use. I assume that problem can be solved through an implementation change in mirclient.

[Daniel van Vugt (vanvugt)]

Do we need to hold 0.23.5 for this?

review: Needs Information

[Andreas Pokorny (andreas-pokorny)]

> I think this is fine as a workaround while the definitions are improved in
> this area.
>
> The physical buffer size is something the client sets, and the onscreen size
> is something server/shell sets.
>
> The legacy problem has to do with how the sizes (physical and on-screen) were
> assumed to be the same. The default stream used to just be resized, and now
> the 1st stream will just be resized.
>
> A function like:
> mir_buffer_stream_set_buffers_size() //sets just the physical buffer size.
> might be what's needed to make things in the non-legacy world a bit nicer and
> clearer.

What is the unit of the physical size? Is this related to the existing mir_buffer_stream_set_scale?

[Andreas Pokorny (andreas-pokorny)]

> Do we need to hold 0.23.5 for this?

We discussed that yesterday. The combination in which resize does not happen is not used by qt, gtk or sdl. So no need to hold 0.23.5 for now.

[Daniel van Vugt (vanvugt)]

Sorry, I don't remember having that discussion. Only asking a few times. Maybe I didn't understand the answer.

review: Abstain

[Alan Griffiths (alan-griffiths)]

The current special treatment of the first buffer seems inconsistent. So, as it is unused in practice, I'd rather not add this code to replicate it.

> The recreate approach has the advantage that it is immediately clear that the
> first buffer you get from the buffer stream will have the right size.
> Additionally the screen update would be tied to the necessary
> mir_surface_apply_spec call. While with the _set_size approach there are
> arguments to have it happen instantly or only after the next buffer swap..
>
> But a current implementation detail makes the recreate approach a bit more
> resource intense since on creation three buffers are allocated. With a
> sequence of resize events two thirds of all created buffers are just allocated
> and freed without use. I assume that problem can be solved through an
> implementation change in mirclient.

My *preference* would be to make the recreate approach (1) work and (2) work efficiently.

But I don't feel strongly about it today.

review: Abstain

[Kevin DuBois (kdub)]

> The current special treatment of the first buffer seems inconsistent. So, as
> it is unused in practice, I'd rather not add this code to replicate it.

Right, it is inconsistent, but it is used, iirc. The surface that gets resized by the server has to respond by resizing the 'default' buffer stream. It seemed better to just keep the auto-resizing around for the 'default/legacy' case, instead saying that the first stream gets resized to the size of the surface (as it is in this MP).

Auto-resizing the 1st stream, as opposed to the only a 'legacy' stream, makes it a bit harder to arrange a scene. Like, if I wanted a 200x200 surface with 4 100x100 windows, and then the server resizes the surface to 500x500, the 1st stream would fill the space after the resize message. The current design is a bit broken for other reasons (no atomic scene arrangements), but I think this might complicate the matter (esp if we're going to use 'tips' for this intented purpose).

This region is due for a bit of cleanup with the atomic updates topic that's coming around soon, maybe we WIP the branch until then, and then roll it into a larger fix to make the scene arrangement and sizes more consistent?

[Andreas Pokorny (andreas-pokorny)]

> > The current special treatment of the first buffer seems inconsistent. So, as
> > it is unused in practice, I'd rather not add this code to replicate it.
>
> Right, it is inconsistent, but it is used, iirc. The surface that gets resized
> by the server has to respond by resizing the 'default' buffer stream. It
> seemed better to just keep the auto-resizing around for the 'default/legacy'
> case, instead saying that the first stream gets resized to the size of the
> surface (as it is in this MP).
>
> Auto-resizing the 1st stream, as opposed to the only a 'legacy' stream, makes
> it a bit harder to arrange a scene. Like, if I wanted a 200x200 surface with 4
> 100x100 windows, and then the server resizes the surface to 500x500, the 1st
> stream would fill the space after the resize message. The current design is a
> bit broken for other reasons (no atomic scene arrangements), but I think this
> might complicate the matter (esp if we're going to use 'tips' for this
> intented purpose).
>
> This region is due for a bit of cleanup with the atomic updates topic that's
> coming around soon, maybe we WIP the branch until then, and then roll it into
> a larger fix to make the scene arrangement and sizes more consistent?

The reason why I proposed this is because we resized the first stream when old buffer semantics were used. But not without another "but". It only did so for the initial configuration - if the users switches to different "first" stream with another spec update it stopped working

We have so many wrong behaviors that I wasnt sure what to expect when changing gtk+.

I also do not like the auto resizing.. So I guess we should focus on making client driven resizes nicer then.

[Daniel van Vugt (vanvugt)]

I don't want to be a blocker here. I didn't realize mir_surface_get_buffer_stream was getting deprecated.

All toolkits want is a:
   mir_surface_swap_buffers(s);
which they currently have to do via:
   mir_buffer_stream_swap_buffers_sync(mir_surface_get_buffer_stream(s));

It's already annoying enough. I don't endorse us making it worse. But that's not the fault of this branch.

review: Abstain

[Andreas Pokorny (andreas-pokorny)]

Rejecting - the fact that OBS with set_streams would resize buffer stream is fishy for the more interesting use cases.

[Andreas Pokorny (andreas-pokorny)]

*would resize the first buffer stream*

Unmerged revisions

3617. By Andreas Pokorny on 2016-08-01

rename new method

3616. By Andreas Pokorny on 2016-07-27

spell out the actual expectation: no exception when attempting to resize a presentation chain

3615. By Andreas Pokorny on 2016-07-27

Tests and additional changes to handle the PresentationChain case

3614. By Andreas Pokorny on 2016-07-27

Make NBS behave like OBS - resize the first buffer stream

With OBS mir::scene::BasicSurface would pull out the first stream as the 'default' stream and forward all resizes to that stream. With NBS this is done by the client library. So if no buffer stream is configured via the mir_surface_spec_add_buffer_stream or mir_surface_spec_set_streams API the client side surface will forward the content of resize events to the 'default' stream.

Since we deprecate using mir_surface_get_buffer_stream users are forces to specify the streams or presentation queues they want for display. But streams buffers lack a resize functionality so switching to the new function would not be a viable option.

=== modified file 'src/client/connection_surface_map.h'
--- src/client/connection_surface_map.h	2016-06-06 11:51:31 +0000
+++ src/client/connection_surface_map.h	2016-08-01 12:51:16 +0000
@@ -39,6 +39,7 @@
     void erase(frontend::SurfaceId surface_id);
 
     void with_stream_do(frontend::BufferStreamId stream_id, std::function<void(ClientBufferStream*)> const& exec) const override;
+    void if_stream_do(frontend::BufferStreamId stream_id, std::function<void(ClientBufferStream*)> const& exec) const override;
     void with_all_streams_do(std::function<void(ClientBufferStream*)> const&) const override;
 
     void insert(frontend::BufferStreamId stream_id, std::shared_ptr<ClientBufferStream> const& chain);

=== modified file 'src/client/mir_connection.cpp'
--- src/client/mir_connection.cpp	2016-06-28 12:18:18 +0000
+++ src/client/mir_connection.cpp	2016-08-01 12:51:16 +0000
@@ -1252,3 +1252,13 @@
     released_buffer->set_buffer_id(buffer->rpc_id());
     server.release_buffers(&request, ignored.get(), gp::NewCallback(ignore));
 }
+
+void MirConnection::resize_stream(mir::frontend::BufferStreamId const id, mir::geometry::Size const& size) const
+{
+    surface_map->if_stream_do(
+        id,
+        [size](mcl::ClientBufferStream* stream)
+        {
+            stream->set_size(size);
+        });
+}

=== modified file 'src/client/mir_connection.h'
--- src/client/mir_connection.h	2016-06-28 12:18:18 +0000
+++ src/client/mir_connection.h	2016-08-01 12:51:16 +0000
@@ -199,6 +199,7 @@
         mir::geometry::Size size, MirPixelFormat format, MirBufferUsage usage,
         mir_buffer_callback callback, void* context);
     void release_buffer(mir::client::MirBuffer* buffer);
+    void resize_stream(mir::frontend::BufferStreamId const id, mir::geometry::Size const& size) const;
 
 private:
     //google cant have callbacks with more than 2 args

=== modified file 'src/client/mir_surface.cpp'
--- src/client/mir_surface.cpp	2016-06-24 07:41:32 +0000
+++ src/client/mir_surface.cpp	2016-08-01 12:51:16 +0000
@@ -45,6 +45,7 @@
 namespace mp = mir::protobuf;
 namespace gp = google::protobuf;
 namespace md = mir::dispatch;
+namespace mf = mir::frontend;
 
 namespace
 {
@@ -153,6 +154,13 @@
             input_platform->create_input_receiver( surface_proto.fd(0), keymapper, handle_event_callback));
     }
 
+    if (spec.streams.is_set() && spec.streams.value().size() >= 1)
+    {
+        auto first_stream = spec.streams.value().front();
+        window_stream_id = mf::BufferStreamId{first_stream.stream_id};
+    }
+
+
     std::lock_guard<decltype(handle_mutex)> lock(handle_mutex);
     valid_surfaces.insert(this);
 }
@@ -456,6 +464,8 @@
         size = { mir_resize_event_get_width(resize_event), mir_resize_event_get_height(resize_event) };
         if (default_stream)
             default_stream->set_size(size);
+        else if (window_stream_id.is_set())
+            connection()->resize_stream(window_stream_id.value(), size);
         break;
     }
     default:
@@ -604,6 +614,8 @@
     if (spec.streams.is_set())
     {
         default_stream = nullptr;
+        if (spec.streams.value().size() >= 1)
+            window_stream_id = mf::BufferStreamId{spec.streams.value().front().stream_id};
         for(auto const& stream : spec.streams.value())
         {
             auto const new_stream = surface_specification->add_stream();

=== modified file 'src/client/mir_surface.h'
--- src/client/mir_surface.h	2016-06-09 00:38:26 +0000
+++ src/client/mir_surface.h	2016-08-01 12:51:16 +0000
@@ -26,6 +26,7 @@
 
 #include "mir/client_platform.h"
 #include "mir/frontend/surface_id.h"
+#include "mir/frontend/buffer_stream_id.h"
 #include "mir/optional_value.h"
 #include "mir/geometry/dimensions.h"
 #include "mir/geometry/displacement.h"
@@ -249,7 +250,7 @@
     MirPixelFormat format;
     MirBufferUsage usage;
     uint32_t output_id;
-
+    mir::optional_value<mir::frontend::BufferStreamId> window_stream_id;
 };
 
 #endif /* MIR_CLIENT_PRIVATE_MIR_WAIT_HANDLE_H_ */

=== modified file 'src/client/surface_map.cpp'
--- src/client/surface_map.cpp	2016-06-06 11:51:31 +0000
+++ src/client/surface_map.cpp	2016-08-01 12:51:16 +0000
@@ -76,6 +76,20 @@
     }
 }
 
+void mcl::ConnectionSurfaceMap::if_stream_do(
+    mf::BufferStreamId stream_id, std::function<void(ClientBufferStream*)> const& exec) const
+{
+    std::shared_lock<decltype(guard)> lk(guard);
+    auto const it = streams.find(stream_id);
+    if (it != streams.end())
+    {
+        auto const stream = it->second;
+        lk.unlock();
+        exec(stream.get());
+    }
+}
+
+
 void mcl::ConnectionSurfaceMap::with_all_streams_do(std::function<void(ClientBufferStream*)> const& fn) const
 {
     std::shared_lock<decltype(guard)> lk(guard);

=== modified file 'src/client/surface_map.h'
--- src/client/surface_map.h	2016-06-06 11:51:31 +0000
+++ src/client/surface_map.h	2016-08-01 12:51:16 +0000
@@ -44,6 +44,8 @@
         frontend::SurfaceId surface_id, std::function<void(MirSurface*)> const& exec) const = 0;
     virtual void with_stream_do(
         frontend::BufferStreamId stream_id, std::function<void(ClientBufferStream*)> const& exec) const = 0;
+    virtual void if_stream_do(
+        frontend::BufferStreamId stream_id, std::function<void(ClientBufferStream*)> const& exec) const = 0;
     virtual void with_all_streams_do(std::function<void(ClientBufferStream*)> const&) const = 0;
     virtual std::shared_ptr<MirBuffer> buffer(int buffer_id) const = 0;
     virtual void insert(int buffer_id, std::shared_ptr<MirBuffer> const& buffer) = 0;

=== modified file 'tests/unit-tests/client/test_client_mir_surface.cpp'
--- tests/unit-tests/client/test_client_mir_surface.cpp	2016-06-08 12:30:14 +0000
+++ tests/unit-tests/client/test_client_mir_surface.cpp	2016-08-01 12:51:16 +0000
@@ -27,6 +27,7 @@
 #include "mir/client_platform.h"
 #include "mir/client_platform_factory.h"
 #include "src/client/mir_surface.h"
+#include "src/client/mir_presentation_chain.h"
 #include "src/client/mir_connection.h"
 #include "src/client/default_connection_configuration.h"
 #include "src/client/rpc/null_rpc_report.h"
@@ -73,6 +74,16 @@
 namespace
 {
 
+struct StubPresentationChain : MirPresentationChain
+{
+    int id;
+    StubPresentationChain(int id) : id{id} {}
+    void submit_buffer(mir::client::MirBuffer*) override  {}
+    MirConnection* connection() const override {return nullptr;}
+    int rpc_id() const override {return id;}
+    char const* error_msg() const override {return nullptr;}
+};
+
 struct MockServerPackageGenerator : public mt::StubServerTool
 {
     MockServerPackageGenerator()
@@ -503,8 +514,12 @@
 TEST_F(MirClientSurfaceTest, resizes_streams_and_calls_callback_if_no_customized_streams)
 {
     using namespace testing;
-    auto mock_stream = std::make_shared<mtd::MockClientBufferStream>(); 
+    auto const stream_id = mir::frontend::BufferStreamId(2);
+    auto mock_stream = std::make_shared<mtd::MockClientBufferStream>();
     auto mock_input_platform = std::make_shared<NiceMock<MockClientInputPlatform>>();
+
+    surface_map->insert(stream_id, mock_stream);
+
     ON_CALL(*mock_input_platform, create_input_receiver(_,_,_))
         .WillByDefault(Return(std::make_shared<mt::TestDispatchable>([]{})));
     ON_CALL(*mock_stream, rpc_id()).WillByDefault(Return(mir::frontend::BufferStreamId(2)));
@@ -516,31 +531,60 @@
     MirSurface surface{connection.get(), *client_comm_channel, nullptr,
         mock_stream, mock_input_platform, spec, surface_proto, wh};
     surface.handle_event(*ev);
-    surface_map->erase(mir::frontend::BufferStreamId(2));
+    surface_map->erase(stream_id);
 }
 
 TEST_F(MirClientSurfaceTest, resizes_streams_and_calls_callback_if_customized_streams)
 {
     using namespace testing;
+    auto const stream_id = mir::frontend::BufferStreamId(2);
     auto mock_stream = std::make_shared<NiceMock<mtd::MockClientBufferStream>>();
     auto mock_input_platform = std::make_shared<NiceMock<MockClientInputPlatform>>();
-    ON_CALL(*mock_stream, rpc_id()).WillByDefault(Return(mir::frontend::BufferStreamId(2)));
+
+    surface_map->insert(stream_id, mock_stream);
+
+    ON_CALL(*mock_stream, rpc_id()).WillByDefault(Return(stream_id));
     ON_CALL(*mock_input_platform, create_input_receiver(_,_,_))
         .WillByDefault(Return(std::make_shared<mt::TestDispatchable>([]{})));
 
     geom::Size size(120, 124);
-    EXPECT_CALL(*mock_stream, set_size(size)).Times(0);
+    EXPECT_CALL(*mock_stream, set_size(size)).Times(1);
     auto ev = mir::events::make_event(mir::frontend::SurfaceId(2), size);
     MirSurface surface{connection.get(), *client_comm_channel, nullptr,
         mock_stream, mock_input_platform, spec, surface_proto, wh};
 
     MirSurfaceSpec spec;
     std::vector<ContentInfo> info =
-        { ContentInfo{ geom::Displacement{0,0}, 2, geom::Size{1,1}} };
+        { ContentInfo{ geom::Displacement{0,0}, stream_id.as_value(), geom::Size{1,1}} };
     spec.streams = info;
     surface.modify(spec)->wait_for_all();
     surface.handle_event(*ev);
-    surface_map->erase(mir::frontend::BufferStreamId(2));
+    surface_map->erase(stream_id);
+}
+
+TEST_F(MirClientSurfaceTest, do_not_resize_presentation_chain)
+{
+    using namespace testing;
+    auto const stream_id = mir::frontend::BufferStreamId(17);
+    auto stub_chain = std::make_shared<StubPresentationChain>(stream_id.as_value());
+    auto mock_input_platform = std::make_shared<NiceMock<MockClientInputPlatform>>();
+
+    surface_map->insert(stream_id, stub_chain);
+
+    ON_CALL(*mock_input_platform, create_input_receiver(_,_,_))
+        .WillByDefault(Return(std::make_shared<mt::TestDispatchable>([]{})));
+
+    geom::Size size(90, 80);
+    auto ev = mir::events::make_event(mir::frontend::SurfaceId(2), size);
+    MirSurface surface{connection.get(), *client_comm_channel, nullptr,
+        nullptr, mock_input_platform, spec, surface_proto, wh};
+
+    MirSurfaceSpec spec;
+    std::vector<ContentInfo> info =
+        { ContentInfo{ geom::Displacement{0,0}, stream_id.as_value(), geom::Size{1,1}} };
+    spec.streams = info;
+    surface.modify(spec)->wait_for_all();
+    EXPECT_NO_THROW(surface.handle_event(*ev););
 }
 
 TEST_F(MirClientSurfaceTest, parameters_are_unhooked_from_stream_sizes)

=== modified file 'tests/unit-tests/client/test_protobuf_rpc_channel.cpp'
--- tests/unit-tests/client/test_protobuf_rpc_channel.cpp	2016-06-24 17:52:00 +0000
+++ tests/unit-tests/client/test_protobuf_rpc_channel.cpp	2016-08-01 12:51:16 +0000
@@ -63,6 +63,8 @@
         void(mir::frontend::SurfaceId, std::function<void(MirSurface*)> const&));
     MOCK_CONST_METHOD2(with_stream_do,
         void(mir::frontend::BufferStreamId, std::function<void(mcl::ClientBufferStream*)> const&));
+    MOCK_CONST_METHOD2(if_stream_do,
+        void(mir::frontend::BufferStreamId, std::function<void(mcl::ClientBufferStream*)> const&));
     MOCK_CONST_METHOD1(with_all_streams_do,
         void(std::function<void(mcl::ClientBufferStream*)> const&));
     MOCK_CONST_METHOD1(buffer, std::shared_ptr<mcl::MirBuffer>(int));
@@ -82,6 +84,10 @@
         mir::frontend::BufferStreamId, std::function<void(mcl::ClientBufferStream*)> const&) const override
     {
     }
+    void if_stream_do(mir::frontend::BufferStreamId,
+                      std::function<void(mcl::ClientBufferStream*)> const&) const override
+    {
+    }
     void with_all_streams_do(std::function<void(mcl::ClientBufferStream*)> const&) const override
     {
     }