652931 - Proper handling of large mark stack insertion

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Gregor Wagner [:gwagner]]

I noticed that for the FOTN the large markstack gets pretty full.

Comment 1

[Gregor Wagner [:gwagner]]

Attachment: patch

Comment 2

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Wait, this seems like a serious bug to me. Is there any guarantee that the assertion will succeed?

If push returns false, we should probably just do the normal thing and not use the large object path.

Thanks for finding this!

Comment 3

[Gregor Wagner [:gwagner]]

Yeah I wanted to ask you why we don't check the return value there.
I will come up with another patch.

Comment 4

[Gregor Wagner [:gwagner]]

Attachment: patch

Comment 5

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Comment on attachment 528863 [details] [diff] [review]
patch

Thanks.

Another way to do it is like this:
  if (...is large dense array... && gcmarker->largeStack.push(LargeMarkItem(obj)))
    /* do thing, push succeeded */;
  else
    clasp->trace(gcmarker, obj);

I'm not sure if this is better or worse, but it avoids code duplication.

Comment 6

[Gregor Wagner [:gwagner]]

Yeah I was thinking of the same code but I have never seen it in SM so I didn't want to start with it :)

Comment 7

[Gregor Wagner [:gwagner]]

http://hg.mozilla.org/tracemonkey/rev/ebc5f8a4b233

Comment 8

[Chris Leary [:cdleary] (not checking bugmail)]

cdleary-bot mozilla-central merge info:
http://hg.mozilla.org/mozilla-central/rev/ebc5f8a4b233

Comment 9

[Robert Sayre]

Do we need this fix on Aurora or Beta?

Comment 10

[Gregor Wagner [:gwagner]]

(In reply to comment #9)
> Do we need this fix on Aurora or Beta?

Depends if bug 616666 is in there.

Comment 12

[Gregor Wagner [:gwagner]]

Bill is the explicit mark stack on Aurora?

Comment 13

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

(In reply to comment #12)
> Bill is the explicit mark stack on Aurora?

It seems like Aurora branched on 4/12 and bug 616666 went in on 4/26. So we're good.

It would be nice if there were a web page that listed the time and changeset id of each branch. Maybe there's an easy way to figure this out, but I ended up just searching through hg.mozilla.org, which is pretty annoying.

Comment 14

[Dão Gottwald [:dao]]

> It would be nice if there were a web page that listed the time and changeset
> id of each branch. Maybe there's an easy way to figure this out, but I ended
> up just searching through hg.mozilla.org, which is pretty annoying.

Just let cdleary-bot set the target milestone in bugzilla?