Apologies for the double-post, lets triage in this separate thread.
MemShrink triage: 2015-12-03
Triage URL: http://mzl.la/1yYeaGL
(Add name and topic to discuss here)
- P1 High importance, will follow up periodically
- P2 Important issue, possibly already being worked on
- P3 Real issue, someone will get to it if they have time
- moreinfo We need logs, or clarifying information
- invalid Misclassified, remove the MemShrink tag
5 bugs to triage
1220488 - Core :: Audio/Video: Playback - FireFox memory leaks and issues, rampant on a particular site
erahm, what do you think? I'm leaning P1. 95% heap-unclassified is
pretty ridiculous, I think we at least need better reporting here.
Additionally it feels like we really shouldn't be using that much memory
for videos that aren't actively playing (this would be particularly bad
1221503 - Firefox :: General - Firefox is slow (Windows swapping all the time)
erahm, what do you think? I'm inclined to close this as WFM, I didn't
see anything that stuck out as egregious, this feels more like a
"Firefox uses more memory than a decade ago" bug.
1224573 - Core :: Audio/Video: Playback - Leak MediaMemoryTracker
erahm, what do you think? I'm inclined to P2 this. It's possible that
the async reporting itself is causing a leak, in that case I might lean
Core :: Hardware Abstraction Layer (HAL) - [e10s] (Win) Nightly
(FF45.0a1) main process use a lot of memory in the last builds
It looks like the reporter is suggesting this is fixed, so probably close WFM.
mconley pointed out that this might be a regression in session restore on e10s from bug 1209689, that points to a rather large talos regression in bug 1227275, which now that I think of it we should probably MemShrink (discussion further down).
1230110 - Core :: DOM - Leak with in <img>
P3, odd usage of html and there's a patch inbound. But it also looks
like we want uplift it, so maybe a higher priority is warranted.
mccr8, what do you think?
- 1227275 -
Firefox :: General - 70-80% e10s sessionrestore regression on fx-team (v.45) on Nov 20, 2015 from push 924d421d766a
Rather large regression, it's noted that this brings us back in line w/
non-e10s, but this is one of the few areas where e10s has been a memory