Kan-Ru Chen [:kanru] (UTC+9) Reporter
	Description • 7 years ago

4,723.50 MB (100.0%) -- explicit
├──2,421.90 MB (51.27%) ── heap-unclassified
├──1,375.52 MB (29.12%) -- js-non-window
│  ├──1,295.88 MB (27.43%) -- zones
│  │  ├──1,270.48 MB (26.90%) -- zone(0x7fb7e21c1000)
│  │  │  ├────863.84 MB (18.29%) ── unused-gc-things
│  │  │  ├────180.09 MB (03.81%) ++ (392 tiny)
│  │  │  ├────100.25 MB (02.12%) -- strings
│  │  │  │    ├───88.66 MB (01.88%) -- malloc-heap
│  │  │  │    │   ├──59.86 MB (01.27%) ── two-byte
│  │  │  │    │   └──28.80 MB (00.61%) ── latin1
│  │  │  │    └───11.59 MB (00.25%) ++ gc-heap
│  │  │  ├─────76.72 MB (01.62%) -- compartment([System Principal], resource://gre/modules/Promise.jsm)
│  │  │  │     ├──67.09 MB (01.42%) ++ classes
│  │  │  │     └───9.63 MB (00.20%) ++ (5 tiny)
│  │  │  └─────49.58 MB (01.05%) ++ compartment([System Principal], resource://gre/modules/Sqlite.jsm)
│  │  └─────25.40 MB (00.54%) ++ (14 tiny)
│  ├─────53.10 MB (01.12%) ++ runtime
│  └─────26.55 MB (00.56%) ++ gc-heap

The large amount of heap-unclassified may come from CycleCollectedJSContext (similar to bug 1333035 comment 2)

If I disable the activity-stream addon then the leak stops.

	Kan-Ru Chen [:kanru] (UTC+9) Reporter
	Comment 2 • 7 years ago

On my desktop I find the about:memory report very similar. Are we leaking Promise.jsm promises?

6,620.73 MB (100.0%) -- explicit
├──5,051.59 MB (76.30%) ── heap-unclassified
├────758.10 MB (11.45%) -- js-non-window
│    ├──682.15 MB (10.30%) -- zones
│    │  ├──671.23 MB (10.14%) -- zone(0x7f603affc000)
│    │  │  ├──257.03 MB (03.88%) ++ (397 tiny)
│    │  │  ├──163.78 MB (02.47%) -- compartment([System Principal], resource://gre/modules/Promise.jsm)
│    │  │  │  ├──139.49 MB (02.11%) ++ classes
│    │  │  │  └───24.29 MB (00.37%) ++ (4 tiny)
│    │  │  ├──153.35 MB (02.32%) ── unused-gc-things
│    │  │  └───97.07 MB (01.47%) -- compartment([System Principal], resource://gre/modules/Sqlite.jsm)
│    │  │      ├──66.22 MB (01.00%) ++ classes
│    │  │      └──30.85 MB (00.47%) ++ (5 tiny)
│    │  └───10.92 MB (00.16%) ++ (8 tiny)
│    └───75.95 MB (01.15%) ++ (2 tiny)

	Eric Rahm [:erahm]
	Comment 3 • 7 years ago

Kan-Ru, can you still repro this?

	Eric Rahm [:erahm]
	Comment 4 • 7 years ago

mossop, you seem to know about activity stream blockers. I'm not sure how you're tracking them, but this should be one.

	Dave Townsend [:mossop]
	Comment 5 • 7 years ago

Tim, are you tracking this?

	Kan-Ru Chen [:kanru] (UTC+9) Reporter
	Comment 6 • 7 years ago

(In reply to Eric Rahm [:erahm] (please no mozreview requests) from comment #3)
> Kan-Ru, can you still repro this?

I haven't seen this for a while. Maybe it's gone?

	Ed Lee :Mardak
	Comment 7 • 7 years ago

(In reply to Kan-Ru Chen [:kanru] (UTC+8) from comment #0)
> If I disable the activity-stream addon then the leak stops.
Sounds like this was analyzed with the test pilot SDK add-on. The system-addon is rewritten as a plain bootstrap add-on enabled on nightly via "browser.newtabpage.activity-stream.enabled"

	Kan-Ru Chen [:kanru] (UTC+9) Reporter
	Comment 8 • 7 years ago

(In reply to Ed Lee :Mardak from comment #7)
> (In reply to Kan-Ru Chen [:kanru] (UTC+8) from comment #0)
> > If I disable the activity-stream addon then the leak stops.
> Sounds like this was analyzed with the test pilot SDK add-on. The
> system-addon is rewritten as a plain bootstrap add-on enabled on nightly via
> "browser.newtabpage.activity-stream.enabled"

I'll try to enable the builtin version and see if I can still reproduce.

	Kan-Ru Chen [:kanru] (UTC+9) Reporter
	Comment 9 • 7 years ago

I haven't seen this for a while. I'm closing this WORKSFORME for now so we don't need to track it.