Erin Lancaster [:elan] Reporter
	Description • 8 years ago

This bug was filed from the Socorro interface and is 
report bp-6fe55baf-4a12-4c2d-9f0a-ab3ea2160312.
=============================================================

	Erin Lancaster [:elan] Reporter
	Comment 1 • 8 years ago

This impacts e10s, :naveed can we get an assignee?

	Andrew McCreight [:mccr8]
	Comment 2 • 8 years ago

This is just an out of memory crash. It looks like Firefox is trying to create a JS runtime for a worker, and it fails.

	Andrew McCreight [:mccr8]
	Comment 3 • 8 years ago

Actually, it looks like there aren't any crashes on versions more recently than 46. Maybe bug 1247122 fixed this (which landed in 47)? Andrea, does that make sense? I don't really understand what that other bug did.

	Andrew McCreight [:mccr8]
	Comment 4 • 8 years ago

Oh I see, it looks like it made the worker creation fail if the runtime creation failed, rather than just crashing. If this is a problem on 46, maybe uplifting that other bug would make sense, at least to get cleaner beta experiment info.

	Kyle Huey (Exited; not receiving bugmail, old account, do not use)
	Comment 5 • 8 years ago

(In reply to Andrew McCreight [:mccr8] from comment #4)
> Oh I see, it looks like it made the worker creation fail if the runtime
> creation failed, rather than just crashing. If this is a problem on 46,
> maybe uplifting that other bug would make sense, at least to get cleaner
> beta experiment info.

Right.  The patch is probably pretty straightforward to uplift.

	Andrea Marchesini [:baku]
	Comment 6 • 8 years ago

Sure. Uplift requested.

	Benjamin Smedberg
	Comment 7 • 8 years ago

I don't think uplift is required. We've filed a bug for the OOM situation more generally with e10s, so I'm going to dup this.

	Andrew McCreight [:mccr8]
	Comment 8 • 8 years ago

Per your criteria in comment 0 of bug 1259512, I think this shouldn't be duped over. This crash happens when we fail to create a new JS runtime, which includes allocating a new GC nursery, which is a contiguous block that is WORKER_DEFAULT_NURSERY_SIZE bytes long. Currently, this is 1MB, qualifying this as an OOM | large.

	Jim Mathies [:jimm]
	Comment 9 • 8 years ago

lets keep this under the oom tracker.

	Andrew McCreight [:mccr8]
	Comment 10 • 8 years ago

I think at this point we're not going to backport anything for it to 46, and later versions are not affected, so I'll mark this WFM.

	Jim Mathies [:jimm]
	Comment 11 • 8 years ago

(In reply to Andrew McCreight [:mccr8] from comment #8)
> Per your criteria in comment 0 of bug 1259512, I think this shouldn't be
> duped over. This crash happens when we fail to create a new JS runtime,
> which includes allocating a new GC nursery, which is a contiguous block that
> is WORKER_DEFAULT_NURSERY_SIZE bytes long. Currently, this is 1MB,
> qualifying this as an OOM | large.

Hey mccr8, how is this not affected by our oom problems? Are you saying that we'll address this in other bugs by reducing memory fragmentation, or did we land something somewhere that helps with this?

	Andrew McCreight [:mccr8]
	Comment 12 • 8 years ago

(In reply to Jim Mathies [:jimm] from comment #11)
> Hey mccr8, how is this not affected by our oom problems? Are you saying that
> we'll address this in other bugs by reducing memory fragmentation, or did we
> land something somewhere that helps with this?

These OOM crashes happen when a page tries to create a new DOM worker and fails. On 46, we crash immediately when this happens. With the patch from bug 1247122, which is on 47 and later, we instead throw an exception and continue running. Thus this particular crash is no longer possible.