RESOLVED FIXED 21924
HashTable internal index is not always deleted 
https://bugs.webkit.org/show_bug.cgi?id=21924
Summary HashTable internal index is not always deleted
Frederic Marmond
Reported 2008-10-28 08:48:01 PDT
HashTable::createTable() allocates new HashEntry array, but HashTable::deleteTable() is not always called, so this array may not be deleted. It can then introduce memory leaks, valgrind gives one of around 800ko on a single gmail visit. My fix is to call the deleteTable() in the HashTable destructor. It is secured and fixes the memory leak.
Attachments
fix a memory leak on HashTable (1.13 KB, patch)
2008-10-28 08:49 PDT, Frederic Marmond 		ap: review-
DetailsFormatted DiffDiff
proposed fix (1.46 KB, patch)
2008-10-28 09:49 PDT, Alexey Proskuryakov 		darin: review+
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Frederic Marmond
Comment 1 2008-10-28 08:49:04 PDT
Created attachment 24714 [details] fix a memory leak on HashTable
Alexey Proskuryakov
Comment 2 2008-10-28 09:15:59 PDT
Comment on attachment 24714 [details] fix a memory leak on HashTable HashTable::deleteTable is called from JSGlobalData destructor to avoid having destructors for static objects, which this patch would introduce. However, the problem is real - in WebCore/bindings/js/JSDOMBinding.cpp, DOMObjectHashTableMap isn't properly cleaned up! This only affects worker threads though.
Alexey Proskuryakov
Comment 3 2008-10-28 09:49:41 PDT
Created attachment 24719 [details] proposed fix Fix the problem in WebCore.
Darin Adler
Comment 4 2008-10-28 09:51:46 PDT
Comment on attachment 24719 [details] proposed fix r=me
Alexey Proskuryakov
Comment 5 2008-10-29 03:35:44 PDT
Committed revision 37968.
Note You need to log in before you can comment on or make changes to this bug.