Open
Bug 1540986
Opened 5 years ago
Updated 2 years ago
Dead zones are not automatically collected
Categories
(Core :: JavaScript: GC, defect, P3)
Core
JavaScript: GC
Tracking
()
NEW
Tracking | Status | |
---|---|---|
firefox68 | --- | affected |
People
(Reporter: jonco, Unassigned)
References
(Blocks 1 open bug)
Details
(Whiteboard: [MemShrink:P2])
The following script causes the shell to allocate 10GB of memory without triggering a collection by allocating memory in many different zones:
for (let i = 0; i < 1000; i++) {
let g = newGlobal({newCompartment: true});
g.eval(`
let size = 10 * 1024 * 1024;
let a = new Int8Array(10 * 1024 * 1024)
for (let i = 0; i < size; i += 4096) {
a[i] = i;
}
`);
}
The memory ends up unreachable but per-zone allocation thresholds are not reached so no GC is triggered.
I think might be what is causing OOMs in our testing. Even if not it suggests we should have some kind of trigger based on the number of zones.
Reporter | ||
Updated•5 years ago
|
Priority: -- → P3
Updated•5 years ago
|
Whiteboard: [MemShrink]
Reporter | ||
Comment 1•5 years ago
|
||
(In reply to Jon Coppeard (:jonco) from comment #0)
I think might be what is causing OOMs in our testing.
I wrote a patch for this and it doesn't help. I now suspect OOMs during reftests are caused by a memory leak.
Updated•5 years ago
|
Whiteboard: [MemShrink] → [MemShrink:P2]
Reporter | ||
Updated•4 years ago
|
Assignee: jcoppeard → nobody
Updated•2 years ago
|
Severity: normal → S3
You need to log in
before you can comment on or make changes to this bug.
Description
•