Mir

Comment 1 for bug 1418002

Revision history for this message
Alexandros Frantzis (afrantzis) wrote :

(Set importance to High since this is a potential CI/autolanding blocker)

Valgrind is known to be pathologically slow in certain busy loop scenarios. It seems that a recent code change in the compositor (probably lp:~vanvugt/mir/fix-buffers_ready_for_compositor) triggered such a scenario.

The increased runtime causes SurfaceSwapBuffers to fail "buffers_swapped.wait_for(std::chrono::seconds{5})" and end the test prematurely. The memory errors are just a side-effect. They are a result of the swap buffers callback trying to access the stack allocated 'buffers_swapped' variable after it has gone out of scope due to the test returning because of the failure.