Merge lp:~vanvugt/qtmir/fix-1495871 into lp:qtmir

[Daniel van Vugt]

Fix fd leak (actually a whole texture+buffer leak) of zombie surfaces
(LP: #1495871)

I'm not completely confident about this. It seems to work, but the fds in question still come and go under normal usage anyway. So the "leak" is hard to detect reliably.

Added to that; every attempt I try to detect GL errors, I see none. But there are other plausible explanations for that...

Then again, if this branch really is working, it might be working for reasons other than those in the comments :)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:378
http://jenkins.qa.ubuntu.com/job/qtmir-ci/438/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/qtmir-vivid-amd64-ci/134/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/qtmir-vivid-armhf-ci/134/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/qtmir-vivid-i386-ci/16/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/qtmir-wily-amd64-ci/171
    SUCCESS: http://jenkins.qa.ubuntu.com/job/qtmir-wily-armhf-ci/171
        deb: http://jenkins.qa.ubuntu.com/job/qtmir-wily-armhf-ci/171/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/qtmir-wily-i386-ci/16

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/qtmir-ci/438/rebuild

review: Needs Fixing (continuous-integration)

[Daniel d'Andrada (dandrader)]

The following branch already fixes that:
https://code.launchpad.net/~dandrader/qtmir/textureCleanUp-lp1499388/+merge/272452

And no, I didn't copy your branch :)

review: Disapprove

[Daniel van Vugt (vanvugt)]

Wow. Nice timing.

Unmerged revisions

378. By Daniel van Vugt on 2015-09-25

Fix fd leak (actually a whole texture leak) of zombie surfaces
(LP: #1495871)

I think. Fingers crossed.

=== modified file 'src/modules/Unity/Application/mirsurfaceitem.cpp'
--- src/modules/Unity/Application/mirsurfaceitem.cpp	2015-09-18 20:11:50 +0000
+++ src/modules/Unity/Application/mirsurfaceitem.cpp	2015-09-25 09:48:00 +0000
@@ -67,10 +67,6 @@
 
     bool smooth;
 
-    void releaseTexture() {
-        t.reset();
-    }
-
     void setTexture(QSharedPointer<QSGTexture> newTexture) {
         t = newTexture;
     }
@@ -586,9 +582,15 @@
         if (!m_surface->isBeingDisplayed() && window()) {
             disconnect(window(), nullptr, m_surface, nullptr);
         }
-        if (m_textureProvider) {
-            m_textureProvider->releaseTexture();
-        }
+
+        // Note we don't release the texture from m_textureProvider here.
+        // Any attempt to release the texture here will silently succeed and
+        // also leak (file descriptors and GPU resources) because the requisite
+        // GL context is missing in the GUI thread. Such a leak is known as
+        // "orphaning" the EGLImage that is used inside the Mir Buffer object;
+        // https://www.khronos.org/registry/egl/extensions/KHR/EGL_KHR_image_base.txt
+        // Instead, we just wait patiently for the render thread to clean up
+        // properly and calls MirSurfaceItem::releaseResources() below.
     }
 
     m_surface = surface;