Blobs from XHR's seem to stick around forever when DevTools is open |
||||||||||||
Issue descriptionOnly go away when the page reloads
,
Mar 21 2017
It will be deleted when the GC runs. I confirmed that with --js-flags="--expose-gc" and chrome://blob-internals.
,
Mar 21 2017
I'm getting this report from +mwills. Mitchell can you confirm whether or not you were seeing this?
,
Mar 21 2017
Yah initially they were being held onto forever because devtools was open. However, even with devtools closed I still have problems because the blobs I download are not gced. I have code that is more or less: download chunk to blob save blob to indexedDB nolonger reference original blob repeat However, the downloads in /tmp eventually build up and the machine runs out of space in that partition. It looks like potentially the blobs simply aren't getting gced (doing things like opening dev tools sometimes causes the storage usage to drop).
,
Mar 21 2017
I did a few more tests. When I run with the --expose-gc flag and call gc() after I remove my reference to the blob the blobs get deleted (space is reclaimed on /tmp and they disappear from chrome://blob-internals/). This seems to indicate that my issue is chrome is not gcing my blobs so blob storage still things there is a reference. Maybe when available space in /tmp becomes low chrome should trigger a GC or something? Although it would be great if there was a way to do it today in stable chrome. Using Blob#close doesn't seem to help at all (not sure if that's been gutted or not).
,
Mar 22 2017
devtools should not retain anything, please attach a snippet of the code for us to investigate.
,
Mar 22 2017
To reproduce: Create a large file called "large": fallocate -l 1GB large Serve "large" and the attached page in chrome with --js-flags="--expose-gc" if you watch the available space in /tmp it will go down and then go back up after about 1GB (after the download completes and gc is called). However, if you open up devtools (and wait for the current request to complete) the available space in /tmp will continue to decline until the partition runs out of space.
,
Apr 18 2017
Sounds like a devtools bug then? Assigning to pfeldman for triage.
,
Apr 18 2017
I'm going to make a blob-side bug for gc importance / accounting for large blobs.
,
Apr 18 2017
Blob-side bug: https://bugs.chromium.org/p/chromium/issues/detail?id=712684
,
Apr 18 2017
Also relates to https://bugs.chromium.org/p/chromium/issues/detail?id=712693
,
Apr 18 2017
,
May 10 2017
,
Dec 1 2017
,
Dec 1 2017
why was this archived?
,
Dec 1 2017
(feel free to change back to archived, just wondering why)
,
Dec 13 2017
,
Oct 12
,
Nov 14
|
||||||||||||
►
Sign in to add a comment |
||||||||||||
Comment 1 by ricea@chromium.org
, Mar 21 2017