1. 26 Oct, 2019 2 commits
    • Evan Stade's avatar
      Remove NotificationService usage in SafeBrowsingService. · 52ae9e88
      Evan Stade authored
      The various services in SafeBrowsingService seem like good candidates
      to be KeyedServices, but their lifetimes don't match up well. Thus,
      simply replace NotificationService notifications with ProfileObserver.
      
      Additionally, fix some bugs in TelemetryService creation/destruction:
      1. If an OTR profile were created and destroyed, the TelemetryService
         for its original profile would have been released, but not destroyed,
         so it would leak on shutdown.
      2. If an OTR profile were never created, the TelemetryService would
         be destroyed when SafeBrowsingService was destroyed, rather than
         in ShutDown(). The destructor of AndroidTelemetryService would then
         access |profile_| after the profile had been destroyed.
      
      Bug: 268984
      Change-Id: I25c8b092c086f9453e41ceaad5eb34ed1d609100
      Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1877149Reviewed-by: default avatarDaniel Rubery <drubery@chromium.org>
      Reviewed-by: default avatarDavid Benjamin <davidben@chromium.org>
      Commit-Queue: Evan Stade <estade@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#709706}
      52ae9e88
    • Eric Karl's avatar
      Revert "DeJelly - Fix SurfaceAggregator's handling of non-merged OOPIF surfaces" · 8766bc4b
      Eric Karl authored
      This reverts commit 8ccc248e.
      
      Reason for revert: This fix was quick / imperfect due to merge constraints. A more complete fix will be landed following this revert.
      
      Original change's description:
      > DeJelly - Fix SurfaceAggregator's handling of non-merged OOPIF surfaces
      > 
      > In cases where a SurfaceDrawQuad's root render pass was not merged into
      > its parent, we did not correctly handle cases where the root render
      > pass represented a subset of the SurfaceDrawQuad's area. This was
      > possible in OOPIF cases where the root render pass was clipped to
      > visible bounds.
      > 
      > This CL adjusts the root render pass so that it always represents the
      > full SurfaceDrawQuad area.
      > 
      > While it's possible to hit this bug without de-jelly, this is much
      > more common with de-jelly, so the change is restricted to reduce merge
      > risk. A  follow-up will implement this change more generally.
      > 
      > Bug: 1016677, 995965
      > Change-Id: If733065277128d53f0b59483f167d177aa777f54
      > Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1873277
      > Commit-Queue: Eric Karl <ericrk@chromium.org>
      > Reviewed-by: Victor Miura <vmiura@chromium.org>
      > Cr-Commit-Position: refs/heads/master@{#708288}
      
      TBR=vmiura@chromium.org,ericrk@chromium.org
      
      # Not skipping CQ checks because original CL landed > 1 day ago.
      
      Bug: 1016677, 995965
      Change-Id: I13b953da9b96bd4a111a393b448359d08be18b38
      Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1879696Reviewed-by: default avatarEric Karl <ericrk@chromium.org>
      Commit-Queue: Eric Karl <ericrk@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#709705}
      8766bc4b
  2. 25 Oct, 2019 38 commits