tree 56462a6c520d1944a6fc87d023848771efaa6a27
parent aca078b5f8f7a0b3afb10f353e435d514ee08075
author Treehugger Robot <treehugger-gerrit@google.com> 1608084902 +0000
committer Gerrit Code Review <noreply-gerritcodereview@google.com> 1608084902 +0000

Update git submodules

* Update frameworks/support from branch 'androidx-master-dev'
  to 6e109e9e6b22704f2b96fe486bc2dce02407ecb5
  - Merge "Prevent OOM in measureCompose runtime benchmarks" into androidx-master-dev
  - Prevent OOM in measureCompose runtime benchmarks
    
    This switches measureCompose to be suspending and manual-time-control in
    order to call advanceUntilIdle() after composition.dispose().  Giving
    the dispatcher an opportunity to run is required to fully clean up
    memory after cancellation of the LaunchedEffect I added in aosp/1435685.
    This is a similar change to the one I did in aosp/1519221 for a
    different set of benchmarks.
    
    Based on my local Wembley testing, this will cause an apparent
    regression in all the affected benchmarks (a large regression in the
    case of the simplest benchmark cases like OneRect), but not to introduce
    additional noise.  So these benchmarks should remain equally useful if
    we accept a one-time discontinuity and compare to the higher baseline.
    
    Bug: 175626893
    Test: Existing tests
    Change-Id: I6f2329bcc281d6ab8c7ece5e10eec43fe0200aec