summaryrefslogtreecommitdiffstats
path: root/gin/run_microtasks_observer.cc
diff options
context:
space:
mode:
authorfalken <falken@chromium.org>2015-05-24 21:35:39 -0700
committerCommit bot <commit-bot@chromium.org>2015-05-25 04:36:10 +0000
commite345e3c8be999296fc968adf09ca02f838ee430f (patch)
tree9f569cecfd9076723670b446d4e69f85a82ef3d6 /gin/run_microtasks_observer.cc
parenta6edaaff54e9b948c64a896edff992c04e63205d (diff)
downloadchromium_src-e345e3c8be999296fc968adf09ca02f838ee430f.zip
chromium_src-e345e3c8be999296fc968adf09ca02f838ee430f.tar.gz
chromium_src-e345e3c8be999296fc968adf09ca02f838ee430f.tar.bz2
Service Worker: Remove the concept of "fail to activate".
As per spec: https://github.com/slightlyoff/ServiceWorker/issues/659 This patch makes SWs activate even if the activate event handler "failed", i.e., it an uncaught exception occurred or the waitUntil promise was rejected. There are some things that could go wrong when trying to activate a worker: the IPC send may fail, the worker may fail to startup (e.g., due to script cache failure), we may time out the activate event. Rather than handling of all these cases, we just say once the browser process attempts to dispatch an activate event, the worker is committed to being activated. I think it's not worth the complexity of retrying the event dispatch or forcefully evicting the worker. Plus, once you've installed a worker that fails to startup, if we want to recover from that, we must due so at a deeper level, not just during activation. This could potentially cause a worker that never received the activate event to be activated, but that should happen rarely. Three-sided patch: [1] Disable layout tests. https://codereview.chromium.org/1139163003/ [2] Chromium change. (this patch) [3] Fix layout tests. https://codereview.chromium.org/1133493004 BUG=480050 Review URL: https://codereview.chromium.org/1126283005 Cr-Commit-Position: refs/heads/master@{#331257}
Diffstat (limited to 'gin/run_microtasks_observer.cc')
0 files changed, 0 insertions, 0 deletions