690933 - Assertion failure: log2 < tl::BitSize<size_t>::result, at jstl.h:223 or crash with memory corruption

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect)

Description

[Christian Holler (:decoder)]

The following test asserts/crashes on mozilla-central revision 1463dc6308a8 (options -m -n), tested on 32 bit:


var fe="vv";
for (i=0; i<24; i++)
  fe += fe;
var fu=new Function(
  fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe,
  fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe, fe,
  "done"
);


Stepping through the assert causes glibc to terminate:

Program received signal SIGABRT, Aborted.
*** glibc detected *** /srv/repos/mozilla-central/js/src/debug32/js: malloc(): memory corruption: 0x084fbcc8 ***
======= Backtrace: =========
[...]

Comment 1

[Christian Holler (:decoder)]

Hm, the original test also asserts, it's

mozilla-central/js/src/tests/js1_5/Function/regress-338121-02.js 

Isn't that test run with the normal test suite?

Comment 2

[Christian Holler (:decoder)]

autoBisect shows this is probably related to the following changeset:

The first bad revision is:
changeset:   77764:4d10127fd106
user:        Chris Leary <[e-mail]>
date:        Thu Sep 22 13:22:30 2011 -0700
summary:     Bug 684039: rewrite JS LIFO allocator, avoids thrashing. (r=luke)

Comment 3

[Daniel Veditz [:dveditz]]

If the bisected regression is correct this should affect Firefox 9 and not Firefox 8.

Comment 4

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

Attachment: Overflow checks.

I think this is only trunk (mozilla10, like on the bisected bug).

The tests were being skipped because of bug 644241 comment 34. (No good way to specify jstest runs as expected-to-OOM.)

Comment 5

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/ae2293392154

Comment 6

[Justin Wood (:Callek)]

https://hg.mozilla.org/mozilla-central/rev/ae2293392154

Comment 7

[u279076]

Can someone who is already set up to reproduce this bug please verify the fix?

Comment 8

[Christian Holler (:decoder)]

Verified fixed on Firefox 10.

Comment 9

[Daniel Veditz [:dveditz]]

No shipping version had this bug, we can unhide it.

Comment 10

[Christian Holler (:decoder)]

Covered by the jit-test in comment 2.