716013 - "Assertion failure: static_cast<Cell *>(thing)->isMarked(),"

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Attachment: stack

f = (function() {
    for (x in [arguments, arguments]) yield(gczeal(4, function(){}))
})
for (i in f()) {}

asserts js debug shell on m-c changeset ae6e1f90b511 without any CLI arguments at Assertion failure: static_cast<Cell *>(thing)->isMarked(),

(not sure if this is totally correct)

autoBisect shows this is probably related to the following changeset:

The first bad revision is:
changeset:   81322:290b3a7329c7
user:        Brian Hackett
date:        Fri Nov 18 12:54:40 2011 -0800
summary:     Move arguments object private data to a reserved slot, bug 703721.

s-s to play safe even though gczeal requires (4, function(){}).

Comment 1

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

See also bug 704258, bug 713226 and bug 714619.

Comment 2

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Replacing gczeal(4, function(){}) with gczeal(2) does not reproduce, so opening up.

I tested this on 32-bit and 64-bit debug shells on Snow Leopard.

Comment 3

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Bill, is this related to work on incremental GC?

Comment 4

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

Yes. Do you have time to look at this, Brian? I suspect that bug 703721 is causing us to miss an incremental barrier.

Comment 5

[Brian Hackett [Laid off!]]

Yeah, I should be able to look at this later this week.

Comment 6

[Brian Hackett [Laid off!]]

Attachment: patch

Arguments and call objects trace through the generator object in any floating frame they are associated with, except when that floating frame has been copied to the stack and is live.  Moving the stack frame out of the object's private data and to a PrivateValue reserved slot skipped the barrier.

Comment 7

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

Comment on attachment 590195 [details] [diff] [review]
patch

Review of attachment 590195 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks Brian. I'm not entirely sure that we should forbid objects with trace hooks from being placed in the nursery, but we should fix this problem regardless.

Comment 8

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

Oh, I forgot, could you fix the typo in the comment ("aboug")?

Comment 9

[Brian Hackett [Laid off!]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/1b5b7d538230

Comment 10

[Marco Bonardo [:mak]]

https://hg.mozilla.org/mozilla-central/rev/1b5b7d538230

Comment 11

[Christian Holler (:decoder)]

A testcase for this bug was automatically identified at js/src/jit-test/tests/basic/bug716013.js.