27874 2009-07-31 04:21:13 -0700 Change malloc to fastMalloc and free to fastFree in Yarr's RegexInterpreter.cpp 2011-06-16 21:50:25 -0700 1 1 1 Unclassified WebKit JavaScriptCore 528+ (Nightly build) PC All RESOLVED INVALID P2 Normal --- 1 zoltan webkit-unassigned barraclough eric oldest_to_newest 136137 0 zoltan 2009-07-31 04:21:13 -0700 Use fastMalloc and fastFree instead of malloc and free in RegexInterpreter.cpp's methods. 136138 1 33870 zoltan 2009-07-31 04:24:22 -0700 Created attachment 33870 proposed patch 136272 2 33870 eric 2009-07-31 15:29:17 -0700 Comment on attachment 33870 proposed patch Clearing review flag on attachment: 33870 Committing to http://svn.webkit.org/repository/webkit/trunk ... M JavaScriptCore/ChangeLog M JavaScriptCore/yarr/RegexInterpreter.cpp Committed r46643 M JavaScriptCore/yarr/RegexInterpreter.cpp M JavaScriptCore/ChangeLog r46643 = 2fe77a59e15a22979daaf74b11f4283d9b4f2186 (trunk) No changes between current HEAD and refs/remotes/trunk Resetting to the latest refs/remotes/trunk http://trac.webkit.org/changeset/46643 136273 3 eric 2009-07-31 15:29:20 -0700 All reviewed patches have been landed. Closing bug. 137129 4 barraclough 2009-08-04 14:36:35 -0700 Unfortunately this change break the v8 tests (as run in the sunspider command-line harness), so reverted for now in r46778. 137132 5 eric 2009-08-04 14:38:32 -0700 Can you define "break"? :) 137136 6 barraclough 2009-08-04 14:47:02 -0700 Long hang spinning the cpu, then a crash. Smells like something going off the deep end of the stack, but I'm afraid I don't really have time to debug this regression further right now. 137262 7 zoltan 2009-08-05 00:17:18 -0700 Weird. Was it on a MAC, wasn't it? 137304 8 eric 2009-08-05 07:19:55 -0700 It sounds like either the conversion to fastMalloc/fastFree was incomplete here, or that by changing over to fastMalloc we've tripped some other subtle bug in the regexp engine. The latter is bad news and requires further investigation. I worry that the current use of malloc/free could just be papering over some underlying bug, given Gavin's above comment. 137384 9 barraclough 2009-08-05 12:45:06 -0700 It might be worth noting that these mallocs and frees will just go away at some point – these will be replaced with a stack-based allocator, though I don't have a timescale for this right now. 137431 10 barraclough 2009-08-05 14:07:10 -0700 (In reply to comment #7) > Weird. Was it on a MAC, wasn't it? This is on Mac, testing the YARR interpreter by setting: #define ENABLE_YARR 1 #define ENABLE_YARR_JIT 0 422410 11 barraclough 2011-06-16 21:50:25 -0700 These allocations are now stack based, and as such this change is no longer valid. 33870 2009-07-31 04:24:22 -0700 2010-06-10 17:57:48 -0700 proposed patch RegexInterpreter.patch text/plain 2604 zoltan SW5kZXg6IEphdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBKYXZhU2NyaXB0 Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDQ2NjI1KQorKysgSmF2YVNjcmlwdENvcmUvQ2hhbmdl TG9nCSh3b3JraW5nIGNvcHkpCkBAIC0xLDMgKzEsMTYgQEAKKzIwMDktMDctMzEgIFpvbHRhbiBI b3J2YXRoICA8aHpvbHRhbkBpbmYudS1zemVnZWQuaHU+CisKKyAgICAgICAgQ2hhbmdlIG1hbGxv YyB0byBmYXN0TWFsbG9jIGFuZCBmcmVlIHRvIGZhc3RGcmVlIGluIFlhcnIncyBSZWdleEludGVy cHJldGVyLmNwcAorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/ aWQ9Mjc4NzQKKworICAgICAgICBVc2UgZmFzdE1hbGxvYyBhbmQgZmFzdEZyZWUgaW5zdGVhZCBv ZiBtYWxsb2MgYW5kIGZyZWUgaW4gUmVnZXhJbnRlcnByZXRlci5jcHAncyBtZXRob2RzLgorCisg ICAgICAgICogeWFyci9SZWdleEludGVycHJldGVyLmNwcDoKKyAgICAgICAgKEpTQzo6WWFycjo6 SW50ZXJwcmV0ZXI6OmFsbG9jRGlzanVuY3Rpb25Db250ZXh0KToKKyAgICAgICAgKEpTQzo6WWFy cjo6SW50ZXJwcmV0ZXI6OmZyZWVEaXNqdW5jdGlvbkNvbnRleHQpOgorICAgICAgICAoSlNDOjpZ YXJyOjpJbnRlcnByZXRlcjo6YWxsb2NQYXJlbnRoZXNlc0Rpc2p1bmN0aW9uQ29udGV4dCk6Cisg ICAgICAgIChKU0M6OllhcnI6OkludGVycHJldGVyOjpmcmVlUGFyZW50aGVzZXNEaXNqdW5jdGlv bkNvbnRleHQpOgorCiAyMDA5LTA3LTMwICBYYW4gTG9wZXogIDx4bG9wZXpAaWdhbGlhLmNvbT4K IAogICAgICAgICBSZXZpZXdlZCBieSBKYW4gQWxvbnpvLgpJbmRleDogSmF2YVNjcmlwdENvcmUv eWFyci9SZWdleEludGVycHJldGVyLmNwcAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBKYXZhU2NyaXB0Q29yZS95 YXJyL1JlZ2V4SW50ZXJwcmV0ZXIuY3BwCShyZXZpc2lvbiA0NjYyNSkKKysrIEphdmFTY3JpcHRD b3JlL3lhcnIvUmVnZXhJbnRlcnByZXRlci5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTEwNCwxMiAr MTA0LDEyIEBAIHB1YmxpYzoKIAogICAgIERpc2p1bmN0aW9uQ29udGV4dCogYWxsb2NEaXNqdW5j dGlvbkNvbnRleHQoQnl0ZURpc2p1bmN0aW9uKiBkaXNqdW5jdGlvbikKICAgICB7Ci0gICAgICAg IHJldHVybiBuZXcobWFsbG9jKHNpemVvZihEaXNqdW5jdGlvbkNvbnRleHQpICsgKGRpc2p1bmN0 aW9uLT5tX2ZyYW1lU2l6ZSAtIDEpICogc2l6ZW9mKHVpbnRwdHJfdCkpKSBEaXNqdW5jdGlvbkNv bnRleHQoKTsKKyAgICAgICAgcmV0dXJuIG5ldyhmYXN0TWFsbG9jKHNpemVvZihEaXNqdW5jdGlv bkNvbnRleHQpICsgKGRpc2p1bmN0aW9uLT5tX2ZyYW1lU2l6ZSAtIDEpICogc2l6ZW9mKHVpbnRw dHJfdCkpKSBEaXNqdW5jdGlvbkNvbnRleHQoKTsKICAgICB9CiAKICAgICB2b2lkIGZyZWVEaXNq dW5jdGlvbkNvbnRleHQoRGlzanVuY3Rpb25Db250ZXh0KiBjb250ZXh0KQogICAgIHsKLSAgICAg ICAgZnJlZShjb250ZXh0KTsKKyAgICAgICAgZmFzdEZyZWUoY29udGV4dCk7CiAgICAgfQogCiAg ICAgc3RydWN0IFBhcmVudGhlc2VzRGlzanVuY3Rpb25Db250ZXh0CkBAIC0xNTAsMTIgKzE1MCwx MiBAQCBwdWJsaWM6CiAKICAgICBQYXJlbnRoZXNlc0Rpc2p1bmN0aW9uQ29udGV4dCogYWxsb2NQ YXJlbnRoZXNlc0Rpc2p1bmN0aW9uQ29udGV4dChCeXRlRGlzanVuY3Rpb24qIGRpc2p1bmN0aW9u LCBpbnQqIG91dHB1dCwgQnl0ZVRlcm0mIHRlcm0pCiAgICAgewotICAgICAgICByZXR1cm4gbmV3 KG1hbGxvYyhzaXplb2YoUGFyZW50aGVzZXNEaXNqdW5jdGlvbkNvbnRleHQpICsgKCgodGVybS5h dG9tLnBhcmVudGhlc2VzRGlzanVuY3Rpb24tPm1fbnVtU3VicGF0dGVybnMgPDwgMSkgLSAxKSAq IHNpemVvZihpbnQpKSArIHNpemVvZihEaXNqdW5jdGlvbkNvbnRleHQpICsgKGRpc2p1bmN0aW9u LT5tX2ZyYW1lU2l6ZSAtIDEpICogc2l6ZW9mKHVpbnRwdHJfdCkpKSBQYXJlbnRoZXNlc0Rpc2p1 bmN0aW9uQ29udGV4dChvdXRwdXQsIHRlcm0pOworICAgICAgICByZXR1cm4gbmV3KGZhc3RNYWxs b2Moc2l6ZW9mKFBhcmVudGhlc2VzRGlzanVuY3Rpb25Db250ZXh0KSArICgoKHRlcm0uYXRvbS5w YXJlbnRoZXNlc0Rpc2p1bmN0aW9uLT5tX251bVN1YnBhdHRlcm5zIDw8IDEpIC0gMSkgKiBzaXpl b2YoaW50KSkgKyBzaXplb2YoRGlzanVuY3Rpb25Db250ZXh0KSArIChkaXNqdW5jdGlvbi0+bV9m cmFtZVNpemUgLSAxKSAqIHNpemVvZih1aW50cHRyX3QpKSkgUGFyZW50aGVzZXNEaXNqdW5jdGlv bkNvbnRleHQob3V0cHV0LCB0ZXJtKTsKICAgICB9CiAKICAgICB2b2lkIGZyZWVQYXJlbnRoZXNl c0Rpc2p1bmN0aW9uQ29udGV4dChQYXJlbnRoZXNlc0Rpc2p1bmN0aW9uQ29udGV4dCogY29udGV4 dCkKICAgICB7Ci0gICAgICAgIGZyZWUoY29udGV4dCk7CisgICAgICAgIGZhc3RGcmVlKGNvbnRl eHQpOwogICAgIH0KIAogICAgIGNsYXNzIElucHV0U3RyZWFtIHsK