586898 - TM: memory leak due to YARR

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Nicholas Nethercote [inactive]]

quietly building debug32
trace-testing debug32 
[405|  0|405]  66% ================================>                  |    8.2s==30265== 4,075 bytes in 1 blocks are definitely lost in loss record 1 of 1
==30265==    at 0x47E1C1C: malloc (vg_replace_malloc.c:195)
==30265==    by 0x8079B50: js_malloc (jsutil.h:193)
==30265==    by 0x8079D48: JS_ArenaAllocate (jsarena.cpp:163)
==30265==    by 0x813FFD6: js::RegExp::execute(JSContext*, JSString*, unsigned int*, bool, js::Value*) (jsregexp.cpp:195)
==30265==    by 0x8142A41: regexp_exec_sub(JSContext*, JSObject*, unsigned int, js::Value*, int, js::Value*) (jsregexp.cpp:923)
==30265==    by 0x8142BE1: regexp_test(JSContext*, unsigned int, js::Value*) (jsregexp.cpp:946)
==30265==    by 0x8237C85: js::Interpret(JSContext*) (jsinterp.cpp:4699)
==30265==    by 0x80DB089: js::Execute(JSContext*, JSObject*, JSScript*, JSStackFrame*, unsigned int, js::Value*) (jsinterp.cpp:886)
==30265==    by 0x80702EC: JS_ExecuteScript (jsapi.cpp:4756)
==30265==    by 0x804BF97: Process(JSContext*, JSObject*, char*, int) (js.cpp:440)
==30265==    by 0x804CCEC: ProcessArgs(JSContext*, JSObject*, char**, int) (js.cpp:786)
==30265==    by 0x80556D9: shell(JSContext*, int, char**, char**) (js.cpp:5047)
==30265== 
[611|  1|612] 100% ==================================================>|   43.7s
FAILURES:
    /home/njn/moz/ws8/js/src/debug32/shell/js -j -e "const platform='linux2'; const libdir='/home/njn/moz/ws8/js/src/trace-test/lib/';" -f /home/njn/moz/ws8/js/src/trace-test/lib/prolog.js -f /home/njn/moz/ws8/js/src/trace-test/tests/basic/testRegExpTest.js

Comment 1

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

Blast, I felt like something was missing. I think we don't free the lazy-released arena chunks. Checking now. This actually is a problem in the fix from bug 586827.

Comment 2

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

Attachment: Unleak YARR regexp arena chunk.

Bzzzt, I'm wrong. Introduced by YARR patch as you said.

Comment 3

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

CCing sayrer... shouldn't push to m/c until this lands.

Comment 4

[Robert Sayre]

Comment on attachment 465587 [details] [diff] [review]
Unleak YARR regexp arena chunk.

trivial patch, r+

Comment 5

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

http://hg.mozilla.org/tracemonkey/rev/9545b8cf201c