Cache Storage API: recreation of the cache bypassing handle retention? |
|||||
Issue descriptionThe repro steps in issue 713738 - 1. Visit our PWA at https://editor.construct.net 2. Wait for "available offline" notification in lower left 3. Open dev tools network panel and record a reload of the page. 4. Look for results for dispatchWorker.js and jobWorker.js. On my machine I see results that look like this: https://www.dropbox.com/s/zdek5si1iig1yko/chrome-sw-worker.png?dl=0 ... appear to show that the script's cache access pattern causes us to recreate the cache repeatedly in the back end, rather than keeping the handle around for 30 seconds. We should verify and minimize the repro.
,
May 11 2017
I didn't fully explain why reverting is okay. The main issue that 2100433003 addressed was that the quota manager, on startup, needs to know the size of every cache on every origin. It used to be that we actually opened all the caches to answer that question, and keeping them all open was bad for resource usage. Now we sync that data to disk at the CacheStorage level and don't need to open the backends.
,
Jan 17 2018
Possibly a duplicate of https://bugs.chromium.org/p/chromium/issues/detail?id=801024 but have not yet repro'ed so still unsure
,
Jan 19 2018
Unassigning from myself, as apparently I'm not working on it. :P
,
Jan 19 2018
Self-assigning while attempting repro
,
May 29 2018
|
|||||
►
Sign in to add a comment |
|||||
Comment 1 by jkarlin@chromium.org
, May 11 2017