22834 2008-12-12 16:38:06 -0800 Mismatched memory free in the new CSSSelectorList 2022-07-13 15:13:52 -0700 1 1 1 Unclassified WebKit CSS 528+ (Nightly build) All All RESOLVED CONFIGURATION CHANGED P2 Normal --- 1 brettw webkit-unassigned bfulgham bnason deanm erikkay louis ppedriana webkit.review.bot zoltan oldest_to_newest 102398 0 brettw 2008-12-12 16:38:06 -0800 CSSParser::createFloatingSelector makes new CSSSelectors using new. These are passed around and eventually put into a CSSSelectorList. This list uses fastFree to free the data in the for loop in CSSSelectorList::adoptSelectorVector. I think the free should instead be delete. 102399 1 25992 brettw 2008-12-12 16:46:10 -0800 Created attachment 25992 Patch I'm not super familiar with this code, so I'm not sure if there are other ways for the selectors to be created. Please check carefully. 102400 2 25992 darin 2008-12-12 17:03:42 -0800 Comment on attachment 25992 Patch r=me 102500 3 25992 brettw 2008-12-14 11:14:25 -0800 Comment on attachment 25992 Patch I now think this patch is wrong. I'll try to come up with another solution. 102571 4 26026 brettw 2008-12-15 11:12:30 -0800 Created attachment 26026 Patch v2 I believe this patch is correct. The previous patch caused double delete problems, which only appeared on Windows for some reason. The problem is that delete also calls the destructor, so when the selector is "moved" in CSSSelectorList, it calls the destructor, then the destructor is called again when the list is freed. This patch uses fastMalloc for the allocation to match the move and delete operations. This is my first ever use of placement new, so check my syntax carefully :) I added a few comments about the allocation method since certain assumptions are made which aren't obvious (causing bugs like this one). 102601 5 eric 2008-12-15 14:14:10 -0800 I r-'d the old patch now that brett says it's wrong (to remove it from the commit queue) 102606 6 26026 darin 2008-12-15 14:53:09 -0800 Comment on attachment 26026 Patch v2 Sadly, this is still wrong, but now it's wrong for the case where parsing failed. In ~CSSParser(), deleteAllValues(m_floatingSelectors) is called, which will call delete on any selectors that are still in the m_floatingSelectors HashSet. So we can't allocate them in a way where delete won't work. Maybe the best design is to add a function to CSSSelector that causes it to drop its data, and call it before calling delete in CSSSelectorList::adoptSelectorVector. It would be enough to just clear out the m_data union. 102687 7 koivisto 2008-12-16 11:40:53 -0800 We have global operator new which does fastMalloc(). It should be safe to fastFree() objects. 102688 8 koivisto 2008-12-16 11:41:59 -0800 Finding better hackery to make purify happy would be good, but not in expense of loosing optimizations. 102812 9 brettw 2008-12-17 13:56:37 -0800 I think Antti is correct and I don't have any brilliant ideas (after discussing with him a bit on IRC) for improving this code without hurting performance. 105658 10 deanm 2009-01-14 05:18:20 -0800 (In reply to comment #7) > We have global operator new which does fastMalloc(). It should be safe to > fastFree() objects. > I don't think this should be relied upon. The global operator new is optional (USE_SYSTEM_MALLOC), and specifically it's important not to rely on an application global new override, when embedders may want to opt out of that, or supply their own. For example, Chromium Linux current doesn't work w/o USE_SYSTEM_MALLOC, because of complicated allocator mismatches (Paul Pedriana seems to be working on this problem). It would be nice if we could override our new / delete to use a different allocator, but current that would mean that when you use new you get one allocator, and when you use fastFree (which will then use free()), you would get another allocator. It would be allow that malloc/free and new/delete use different incompatible allocators. In general I think new / delete / free mismatches can be a lot of trouble, and it's really important to keep them consistent, our you're really limiting the allocation options for embedders. It's more than just pacifying purify, and we're already our Linux port is already seeing problems with the WebKit allocation design. 105693 11 darin 2009-01-14 10:01:19 -0800 (In reply to comment #10) > (In reply to comment #7) > > We have global operator new which does fastMalloc(). It should be safe to > > fastFree() objects. > > I don't think this should be relied upon. I agree. 105939 12 26766 deanm 2009-01-15 12:06:04 -0800 Created attachment 26766 Patch attempt. This is really not the cleanest approach, and technically by the C++ standard the behavior is undefined. However, I think it's an improvement over the current allocation mismatch. 105947 13 26766 darin 2009-01-15 12:59:29 -0800 Comment on attachment 26766 Patch attempt. I don't think this is enough of an improvement. If we're going to go to this trouble I wish we could do it right. The best way I can think of would be to implement swap for CSSSelector and use swap and delete instead of memcpy and fastFree. 105951 14 deanm 2009-01-15 13:06:46 -0800 Another hacky and probably inappropriate solution would be to zero the old selector after we call memcpy on it, and before calling delete on it. We are still going to need the second half of the patch to control whether we use fastMalloc/fastFree, or delete. However, with the way it's written, it seems like it will be a bit awkward. Any thoughts on zeroing? I will work on the swap implementation. (In reply to comment #13) > (From update of attachment 26766 [review]) > I don't think this is enough of an improvement. > > If we're going to go to this trouble I wish we could do it right. The best way > I can think of would be to implement swap for CSSSelector and use swap and > delete instead of memcpy and fastFree. > 105952 15 brettw 2009-01-15 13:10:45 -0800 I personally like the swap approach. I use it in my code from time to time in similar circumstances. 105954 16 deanm 2009-01-15 13:15:02 -0800 Well, we will be swapping uninitialized memory into the old Selector, and then it's destructor is going to crash. You can't really implement a swap, you would need to implement a copyToAndClearSelf(CSSSelector otherSelector). It seems a bit awkward to me. The other option is to actually call the constructor, and use new[] instead of fastMalloc. But I thought the whole argument originally was that there was some performance at stake, and now it would seem like we're doing a lot more work. Let me know what you think is the best option. Thanks. (In reply to comment #15) > I personally like the swap approach. I use it in my code from time to time in > similar circumstances. > 107172 17 deanm 2009-01-27 02:18:43 -0800 Hi, My patch to use a delete reinterpret_cast<> is effectively the same thing as what's happening now, but it also fixes a real bug. I feel like this is one of those cases where the patch isn't perfect, so instead we'll do nothing. If someone is more familiar with the code and performance requirements, and is going to make a cleaner fix, that would be great. In the meantime my patch would at least fix a bug. (In reply to comment #16) > Well, we will be swapping uninitialized memory into the old Selector, and then > it's destructor is going to crash. You can't really implement a swap, you > would need to implement a copyToAndClearSelf(CSSSelector otherSelector). It > seems a bit awkward to me. > > The other option is to actually call the constructor, and use new[] instead of > fastMalloc. But I thought the whole argument originally was that there was > some performance at stake, and now it would seem like we're doing a lot more > work. > > Let me know what you think is the best option. Thanks. > > (In reply to comment #15) > > I personally like the swap approach. I use it in my code from time to time in > > similar circumstances. > > > 111549 18 tony 2009-02-27 14:12:04 -0800 *** Bug 23473 has been marked as a duplicate of this bug. *** 114547 19 26766 darin 2009-03-20 12:58:46 -0700 Comment on attachment 26766 Patch attempt. > - fastFree(selectorVector[i]); > + // We want to free the memory (which was allocated with new), but we > + // don't want the destructor to run since it will affect the copy > + // we've just made. In theory this is undefined, but operator delete > + // is only defined taking a void*, so in practice it should be ok. > + delete reinterpret_cast<char*>(selectorVector[i]); While this is no better than fastFree in practice, you've convinced me it’s worthwhile because it makes valgrind stop complaining, and this makes it slightly more likely the code would work if we made delete and fastFree incompatible with each other. Once this is done, we still have a bug, but I suppose it’s a less serious one than what we have now. > + // We had two cases in adoptSelectVector. The fast case of a 1 element > + // vector took the CSSSelector directly, which was allocated with new. > + // The second case we allocated a new fastMalloc buffer, which sholud be > + // freed with fastFree, and the destructors called manually. Whoever lands this should fix the typo in "sholud". We also normally do one space after a period, not two. r=me 114565 20 levin 2009-03-20 15:06:50 -0700 Assigned to levin for landing. 114572 21 levin 2009-03-20 15:23:59 -0700 Please add a link to the bug in the changelog in the future. 114602 22 levin 2009-03-20 16:55:33 -0700 Committed as r41877. 179810 23 zoltan 2010-01-12 06:17:07 -0800 FastMalloc mismatch validation system (Platform.h, FAST_FAST_MALLOC_MATCH_VALIDATION=1) crashes on Linux-Qt in CSSSelectorList due to mismatch. Benbuck Nason has a solution for the problem, I reopen the bug. 179832 24 46375 bnason 2010-01-12 07:51:15 -0800 Created attachment 46375 Patch for comment #23 179833 25 webkit.review.bot 2010-01-12 07:51:56 -0800 Attachment 46375 did not pass style-queue: Failed to run "WebKitTools/Scripts/check-webkit-style" exit_code: 1 WebCore/ChangeLog:5: Line contains tab character. [whitespace/tab] [5] WebCore/ChangeLog:6: Line contains tab character. [whitespace/tab] [5] WebCore/ChangeLog:7: Line contains tab character. [whitespace/tab] [5] WebCore/ChangeLog:8: Line contains tab character. [whitespace/tab] [5] WebCore/ChangeLog:9: Line contains tab character. [whitespace/tab] [5] WebCore/ChangeLog:10: Line contains tab character. [whitespace/tab] [5] Total errors found: 6 179834 26 46376 bnason 2010-01-12 07:55:09 -0800 Created attachment 46376 Patch for comment #23 (retry) 180258 27 46376 darin 2010-01-13 09:40:46 -0800 Comment on attachment 46376 Patch for comment #23 (retry) The right fix is to change CSSParser.cpp to call "new CSSSelector" instead of "fastNew<CSSSelector>()". > - delete m_data.m_tagHistory; > + fastDelete(m_data.m_tagHistory); This isn't right. When m_hasRareData is true, then the tag history will be deleted as part of the destruction of OwnPtr. This will call delete, not fastDelete. So changing the code in ~CSSSelector to call fastDelete will not fix the mismatch. But since CSSSelector inherits from Noncopyable, there is no reason to explicitly use fastNew on a CSSSelector -- Noncopyable inherits from FastAllocBase, and so new should automatically use fastNew. This code change seems both incorrect and unnecessary. > - delete s; > + fastDelete(s); Same comment here. Changing these call sites is incompatible with using OwnPtr<CSSSelector> elsewhere. 180635 28 zoltan 2010-01-14 02:32:37 -0800 (In reply to comment #27) > (From update of attachment 46376 [details]) > The right fix is to change CSSParser.cpp to call "new CSSSelector" instead of > "fastNew<CSSSelector>()". We put fastNew into CSSParser.cpp to avoid valgrind's mismatched warnings in bug #25930. 180870 29 26766 eric 2010-01-14 12:58:34 -0800 Comment on attachment 26766 Patch attempt. Clearing Darin Adler's review on this old patch and marking obsolete. This patch was already landed by Levin, see above. 180871 30 26026 eric 2010-01-14 12:58:59 -0800 Comment on attachment 26026 Patch v2 Marking this old patch obsolete now that this bug has been re-opened for a second change. 206789 31 levin 2010-03-31 16:13:42 -0700 Cleared myself from being the assignee (since I only did this for landing the patch). 1883395 32 bfulgham 2022-07-13 15:13:52 -0700 This code has been significantly refactored since this patch was proposed. There doesn't seem to be any action we can take here. 25992 2008-12-12 16:46:10 -0800 2008-12-15 14:12:14 -0800 Patch delete.diff text/plain 1412 brettw ZGlmZiAtLWdpdCBhL1dlYkNvcmUvQ2hhbmdlTG9nIGIvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXgg YjY5YzY0Zi4uNzVlZTIzNiAxMDA2NDQKLS0tIGEvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvV2Vi Q29yZS9DaGFuZ2VMb2cKQEAgLTEsMyArMSwxNSBAQAorMjAwOC0xMi0xMiAgQnJldHQgV2lsc29u ICA8YnJldHR3QGNocm9taXVtLm9yZz4KKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9P UFMhKS4KKworICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9 MjI4MzQKKworICAgICAgICBNYWtlIHRoZSBDU1NTZWxlY3RvciBmcmVlIGluIENTU1NlbGVjdG9y TGlzdCBtYXRjaCB0aGUgd2F5IHRoZQorICAgICAgICBzZWxlY3RvcnMgYXJlIGFsbG9jYXRlZCBp biBDU1NQYXJzZXIgKHZpYSBuZXcgcmF0aGVyIHRoYW4gZmFzdE1hbGxvYykuCisKKyAgICAgICAg KiBjc3MvQ1NTU2VsZWN0b3JMaXN0LmNwcDoKKyAgICAgICAgKFdlYkNvcmU6OkNTU1NlbGVjdG9y TGlzdDo6YWRvcHRTZWxlY3RvclZlY3Rvcik6CisKIDIwMDgtMTItMTEgIFN0ZXZlIEZhbGtlbmJ1 cmcgIDxzZmFsa2VuQGFwcGxlLmNvbT4KIAogICAgICAgICBEZWxldGUgdGhlIHByZXZpb3VzIHRp bWVyLXF1ZXVlIHRpbWVyIGluIHRoZSBtYWluIHRocmVhZCwganVzdCBwcmlvciB0byBzY2hlZHVs aW5nIGEgbmV3IHRpbWVyLgpkaWZmIC0tZ2l0IGEvV2ViQ29yZS9jc3MvQ1NTU2VsZWN0b3JMaXN0 LmNwcCBiL1dlYkNvcmUvY3NzL0NTU1NlbGVjdG9yTGlzdC5jcHAKaW5kZXggNzI3NjE0MS4uZjk2 YTJlMCAxMDA2NDQKLS0tIGEvV2ViQ29yZS9jc3MvQ1NTU2VsZWN0b3JMaXN0LmNwcAorKysgYi9X ZWJDb3JlL2Nzcy9DU1NTZWxlY3Rvckxpc3QuY3BwCkBAIC01NCw3ICs1NCw3IEBAIHZvaWQgQ1NT U2VsZWN0b3JMaXN0OjphZG9wdFNlbGVjdG9yVmVjdG9yKFZlY3RvcjxDU1NTZWxlY3Rvcio+JiBz ZWxlY3RvclZlY3RvcikKICAgICBtX3NlbGVjdG9yQXJyYXkgPSByZWludGVycHJldF9jYXN0PENT U1NlbGVjdG9yKj4oZmFzdE1hbGxvYyhzaXplb2YoQ1NTU2VsZWN0b3IpICogc2VsZWN0b3JWZWN0 b3Iuc2l6ZSgpKSk7CiAgICAgZm9yIChzaXplX3QgaSA9IDA7IGkgPCBzaXplOyArK2kpIHsKICAg ICAgICAgbWVtY3B5KCZtX3NlbGVjdG9yQXJyYXlbaV0sIHNlbGVjdG9yVmVjdG9yW2ldLCBzaXpl b2YoQ1NTU2VsZWN0b3IpKTsKLSAgICAgICAgZmFzdEZyZWUoc2VsZWN0b3JWZWN0b3JbaV0pOwor ICAgICAgICBkZWxldGUgc2VsZWN0b3JWZWN0b3JbaV07CiAgICAgICAgIEFTU0VSVCghbV9zZWxl Y3RvckFycmF5W2ldLmlzTGFzdEluU2VsZWN0b3JMaXN0KCkpOwogICAgIH0KICAgICBtX3NlbGVj dG9yQXJyYXlbc2l6ZSAtIDFdLnNldExhc3RJblNlbGVjdG9yTGlzdCgpOwo= 26026 2008-12-15 11:12:30 -0800 2010-01-14 12:58:59 -0800 Patch v2 selectors.diff text/plain 2050 brettw ZGlmZiAtLWdpdCBhL1dlYkNvcmUvQ2hhbmdlTG9nIGIvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXgg Y2ZjNThjNC4uMzQzMGVjYyAxMDA2NDQKLS0tIGEvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvV2Vi Q29yZS9DaGFuZ2VMb2cKQEAgLTEsMyArMSwxNyBAQAorMjAwOC0xMi0xNSAgQnJldHQgV2lsc29u ICA8YnJldHR3QGNocm9taXVtLm9yZz4KKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9P UFMhKS4KKworICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9 MjI4MzQKKworICAgICAgICBNYWtlIHRoZSBjcmFldGlvbiBvZiBDU1NTZWxlY3RvcnMgbWF0Y2gg dGhlIHdheSB0aGV5IGFyZSBkZWxldGVkCisgICAgICAgIHZpYSBmYXN0TWFsbG9jL2ZyZWUgaW4g Q1NTU2VsZWN0b3JMaXN0LgorCisgICAgICAgICogY3NzL0NTU1BhcnNlci5jcHA6CisgICAgICAg IChXZWJDb3JlOjpDU1NQYXJzZXI6OmNyZWF0ZUZsb2F0aW5nU2VsZWN0b3IpOgorICAgICAgICAq IGNzcy9DU1NQYXJzZXIuaDoKKyAgICAgICAgKiBjc3MvQ1NTU2VsZWN0b3JMaXN0Lmg6CisKIDIw MDgtMTItMTUgIEFudHRpIEtvaXZpc3RvICA8YW50dGlAYXBwbGUuY29tPgogCiAgICAgICAgIFJl dmlld2VkIGJ5IERhcmluIEFkbGVyLgpkaWZmIC0tZ2l0IGEvV2ViQ29yZS9jc3MvQ1NTUGFyc2Vy LmNwcCBiL1dlYkNvcmUvY3NzL0NTU1BhcnNlci5jcHAKaW5kZXggMjFiOGFmMC4uYTNiZTY4MiAx MDA2NDQKLS0tIGEvV2ViQ29yZS9jc3MvQ1NTUGFyc2VyLmNwcAorKysgYi9XZWJDb3JlL2Nzcy9D U1NQYXJzZXIuY3BwCkBAIC00MzcyLDcgKzQzNzIsNyBAQCBVQ2hhciogQ1NTUGFyc2VyOjp0ZXh0 KGludCAqbGVuZ3RoKQogCiBDU1NTZWxlY3RvciogQ1NTUGFyc2VyOjpjcmVhdGVGbG9hdGluZ1Nl bGVjdG9yKCkKIHsKLSAgICBDU1NTZWxlY3Rvciogc2VsZWN0b3IgPSBuZXcgQ1NTU2VsZWN0b3I7 CisgICAgQ1NTU2VsZWN0b3IqIHNlbGVjdG9yID0gbmV3IChmYXN0TWFsbG9jKHNpemVvZihDU1NT ZWxlY3RvcikpKSBDU1NTZWxlY3RvcigpOwogICAgIG1fZmxvYXRpbmdTZWxlY3RvcnMuYWRkKHNl bGVjdG9yKTsKICAgICByZXR1cm4gc2VsZWN0b3I7CiB9CmRpZmYgLS1naXQgYS9XZWJDb3JlL2Nz cy9DU1NQYXJzZXIuaCBiL1dlYkNvcmUvY3NzL0NTU1BhcnNlci5oCmluZGV4IDgzYWFmYTQuLmZi ZDA4OWQgMTAwNjQ0Ci0tLSBhL1dlYkNvcmUvY3NzL0NTU1BhcnNlci5oCisrKyBiL1dlYkNvcmUv Y3NzL0NTU1BhcnNlci5oCkBAIC0xNTIsNiArMTUyLDcgQEAgbmFtZXNwYWNlIFdlYkNvcmUgewog CiAgICAgICAgIGludCB5eXBhcnNlKCk7CiAKKyAgICAgICAgLy8gVGhlIG5ldyBzZWxlY3RvciB3 aWxsIGJlIGFsbG9jYXRlZCB3aXRoIGZhc3RNYWxsb2MuCiAgICAgICAgIENTU1NlbGVjdG9yKiBj cmVhdGVGbG9hdGluZ1NlbGVjdG9yKCk7CiAgICAgICAgIENTU1NlbGVjdG9yKiBzaW5rRmxvYXRp bmdTZWxlY3RvcihDU1NTZWxlY3RvciopOwogCmRpZmYgLS1naXQgYS9XZWJDb3JlL2Nzcy9DU1NT ZWxlY3Rvckxpc3QuaCBiL1dlYkNvcmUvY3NzL0NTU1NlbGVjdG9yTGlzdC5oCmluZGV4IDdhNDFm Y2YuLmRlMTI0MzcgMTAwNjQ0Ci0tLSBhL1dlYkNvcmUvY3NzL0NTU1NlbGVjdG9yTGlzdC5oCisr KyBiL1dlYkNvcmUvY3NzL0NTU1NlbGVjdG9yTGlzdC5oCkBAIC0zNyw2ICszNyw4IEBAIG5hbWVz cGFjZSBXZWJDb3JlIHsKICAgICAgICAgfkNTU1NlbGVjdG9yTGlzdCgpOwogCiAgICAgICAgIHZv aWQgYWRvcHQoQ1NTU2VsZWN0b3JMaXN0JiBsaXN0KTsKKworICAgICAgICAvLyBUaGUgaW5kaXZp ZHVhbCBzZWxlY3RvcnMgbXVzdCBoYXZlIGJlZW4gYWxsb2NhdGVkIHdpdGggZmFzdE1hbGxvYy4K ICAgICAgICAgdm9pZCBhZG9wdFNlbGVjdG9yVmVjdG9yKFZlY3RvcjxDU1NTZWxlY3Rvcio+JiBz ZWxlY3RvclZlY3Rvcik7CiAgICAgICAgIAogICAgICAgICBDU1NTZWxlY3RvciogZmlyc3QoKSBj b25zdCB7IHJldHVybiBtX3NlbGVjdG9yQXJyYXkgPyBtX3NlbGVjdG9yQXJyYXkgOiAwOyB9Cg== 26766 2009-01-15 12:06:04 -0800 2010-01-14 12:58:33 -0800 Patch attempt. x.diff text/plain 2540 deanm ZGlmZiAtLWdpdCBhL1dlYkNvcmUvQ2hhbmdlTG9nIGIvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXgg NGNiZTdhOC4uNDg5YjU5YiAxMDA2NDQKLS0tIGEvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvV2Vi Q29yZS9DaGFuZ2VMb2cKQEAgLTEsMyArMSwxMyBAQAorMjAwOS0wMS0xNSAgRGVhbiBNY05hbWVl ICA8ZGVhbm1AY2hyb21pdW0ub3JnPgorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09Q UyEpLgorCisgICAgICAgIE1ha2Ugc3VyZSB0byBjb25zaXN0ZW50bHkgbWF0Y2ggbmV3L2RlbGV0 ZSBhbmQgZmFzdE1hbGxvYy9mYXN0RnJlZS4KKworICAgICAgICAqIGNzcy9DU1NTZWxlY3Rvckxp c3QuY3BwOgorICAgICAgICAoV2ViQ29yZTo6Q1NTU2VsZWN0b3JMaXN0OjphZG9wdFNlbGVjdG9y VmVjdG9yKToKKyAgICAgICAgKFdlYkNvcmU6OkNTU1NlbGVjdG9yTGlzdDo6ZGVsZXRlU2VsZWN0 b3JzKToKKwogMjAwOS0wMS0xNSAgQ2hyaXMgTWFycmluICA8Y21hcnJpbkBhcHBsZS5jb20+CiAK ICAgICAgICAgUmV2aWV3ZWQgYnkgRGFuIEJlcm5zdGVpbi4KZGlmZiAtLWdpdCBhL1dlYkNvcmUv Y3NzL0NTU1NlbGVjdG9yTGlzdC5jcHAgYi9XZWJDb3JlL2Nzcy9DU1NTZWxlY3Rvckxpc3QuY3Bw CmluZGV4IDcyNzYxNDEuLjc1ZDZmMDQgMTAwNjQ0Ci0tLSBhL1dlYkNvcmUvY3NzL0NTU1NlbGVj dG9yTGlzdC5jcHAKKysrIGIvV2ViQ29yZS9jc3MvQ1NTU2VsZWN0b3JMaXN0LmNwcApAQCAtNTQs NyArNTQsMTEgQEAgdm9pZCBDU1NTZWxlY3Rvckxpc3Q6OmFkb3B0U2VsZWN0b3JWZWN0b3IoVmVj dG9yPENTU1NlbGVjdG9yKj4mIHNlbGVjdG9yVmVjdG9yKQogICAgIG1fc2VsZWN0b3JBcnJheSA9 IHJlaW50ZXJwcmV0X2Nhc3Q8Q1NTU2VsZWN0b3IqPihmYXN0TWFsbG9jKHNpemVvZihDU1NTZWxl Y3RvcikgKiBzZWxlY3RvclZlY3Rvci5zaXplKCkpKTsKICAgICBmb3IgKHNpemVfdCBpID0gMDsg aSA8IHNpemU7ICsraSkgewogICAgICAgICBtZW1jcHkoJm1fc2VsZWN0b3JBcnJheVtpXSwgc2Vs ZWN0b3JWZWN0b3JbaV0sIHNpemVvZihDU1NTZWxlY3RvcikpOwotICAgICAgICBmYXN0RnJlZShz ZWxlY3RvclZlY3RvcltpXSk7CisgICAgICAgIC8vIFdlIHdhbnQgdG8gZnJlZSB0aGUgbWVtb3J5 ICh3aGljaCB3YXMgYWxsb2NhdGVkIHdpdGggbmV3KSwgYnV0IHdlCisgICAgICAgIC8vIGRvbid0 IHdhbnQgdGhlIGRlc3RydWN0b3IgdG8gcnVuIHNpbmNlIGl0IHdpbGwgYWZmZWN0IHRoZSBjb3B5 CisgICAgICAgIC8vIHdlJ3ZlIGp1c3QgbWFkZS4gIEluIHRoZW9yeSB0aGlzIGlzIHVuZGVmaW5l ZCwgYnV0IG9wZXJhdG9yIGRlbGV0ZQorICAgICAgICAvLyBpcyBvbmx5IGRlZmluZWQgdGFraW5n IGEgdm9pZCosIHNvIGluIHByYWN0aWNlIGl0IHNob3VsZCBiZSBvay4KKyAgICAgICAgZGVsZXRl IHJlaW50ZXJwcmV0X2Nhc3Q8Y2hhcio+KHNlbGVjdG9yVmVjdG9yW2ldKTsKICAgICAgICAgQVNT RVJUKCFtX3NlbGVjdG9yQXJyYXlbaV0uaXNMYXN0SW5TZWxlY3Rvckxpc3QoKSk7CiAgICAgfQog ICAgIG1fc2VsZWN0b3JBcnJheVtzaXplIC0gMV0uc2V0TGFzdEluU2VsZWN0b3JMaXN0KCk7CkBA IC02NSwxNSArNjksMjUgQEAgdm9pZCBDU1NTZWxlY3Rvckxpc3Q6OmRlbGV0ZVNlbGVjdG9ycygp CiB7CiAgICAgaWYgKCFtX3NlbGVjdG9yQXJyYXkpCiAgICAgICAgIHJldHVybjsKKworICAgIC8v IFdlIGhhZCB0d28gY2FzZXMgaW4gYWRvcHRTZWxlY3RWZWN0b3IuICBUaGUgZmFzdCBjYXNlIG9m IGEgMSBlbGVtZW50CisgICAgLy8gdmVjdG9yIHRvb2sgdGhlIENTU1NlbGVjdG9yIGRpcmVjdGx5 LCB3aGljaCB3YXMgYWxsb2NhdGVkIHdpdGggbmV3LgorICAgIC8vIFRoZSBzZWNvbmQgY2FzZSB3 ZSBhbGxvY2F0ZWQgYSBuZXcgZmFzdE1hbGxvYyBidWZmZXIsIHdoaWNoIHNob2x1ZCBiZQorICAg IC8vIGZyZWVkIHdpdGggZmFzdEZyZWUsIGFuZCB0aGUgZGVzdHJ1Y3RvcnMgY2FsbGVkIG1hbnVh bGx5LgogICAgIENTU1NlbGVjdG9yKiBzID0gbV9zZWxlY3RvckFycmF5OwotICAgIHdoaWxlICgx KSB7Ci0gICAgICAgIGJvb2wgZG9uZSA9IHMtPmlzTGFzdEluU2VsZWN0b3JMaXN0KCk7Ci0gICAg ICAgIHMtPn5DU1NTZWxlY3RvcigpOwotICAgICAgICBpZiAoZG9uZSkKLSAgICAgICAgICAgIGJy ZWFrOwotICAgICAgICArK3M7CisgICAgYm9vbCBkb25lID0gcy0+aXNMYXN0SW5TZWxlY3Rvckxp c3QoKTsKKyAgICBpZiAoZG9uZSkgeworICAgICAgICBkZWxldGUgczsKKyAgICB9IGVsc2Ugewor ICAgICAgICB3aGlsZSAoMSkgeworICAgICAgICAgICAgcy0+fkNTU1NlbGVjdG9yKCk7CisgICAg ICAgICAgICBpZiAoZG9uZSkKKyAgICAgICAgICAgICAgICBicmVhazsKKyAgICAgICAgICAgICsr czsKKyAgICAgICAgICAgIGRvbmUgPSBzLT5pc0xhc3RJblNlbGVjdG9yTGlzdCgpOworICAgICAg ICB9CisgICAgICAgIGZhc3RGcmVlKG1fc2VsZWN0b3JBcnJheSk7IAogICAgIH0KLSAgICBmYXN0 RnJlZShtX3NlbGVjdG9yQXJyYXkpOyAKIH0KIAogfQo= 46375 2010-01-12 07:51:15 -0800 2010-01-12 07:55:09 -0800 Patch for comment #23 22834.patch text/plain 1738 bnason SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiA1MzE0MCkKKysrIFdlYkNvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBA IC0xLDMgKzEsMTkgQEAKKzIwMTAtMDEtMTIgIEJlbmJ1Y2sgTmFzb24gIDxibmFzb25AbmV0Zmxp eC5jb20+CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKwkJaHR0cHM6 Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTIyODM0CisJCU1pc21hdGNoZWQgbWVt b3J5IGZyZWUgaW4gQ1NTU2VsZWN0b3JMaXN0CisJCQorCQlGaXggbWlzbWF0Y2hlZCBmYXN0TWFs bG9jKCkvZGVsZXRlIGluIENTU1NlbGVjdG9yTGlzdDo6ZGVsZXRlU2VsZWN0b3JzKCkKKwkJYW5k IENTU1NlbGVjdG9yOjp+Q1NTU2VsZWN0b3IoKSB3aGVuIEZBU1RfRkFTVF9NQUxMT0NfTUFUQ0hf VkFMSURBVElPTj0xCisJCWlzIHNldC4KKworICAgICAgICAqIGNzcy9DU1NTZWxlY3Rvci5oOgor ICAgICAgICAoV2ViQ29yZTo6Q1NTU2VsZWN0b3I6On5DU1NTZWxlY3Rvcik6CisgICAgICAgICog Y3NzL0NTU1NlbGVjdG9yTGlzdC5jcHA6CisgICAgICAgIChXZWJDb3JlOjpDU1NTZWxlY3Rvckxp c3Q6OmRlbGV0ZVNlbGVjdG9ycyk6CisKIDIwMTAtMDEtMTIgIEFkYW0gUm9iZW4gIDxhcm9iZW5A YXBwbGUuY29tPgogCiAgICAgICAgIFdpbmRvd3MgYWNjZWxlcmF0ZWQgY29tcG9zaXRpbmcgYnVp bGQgZml4IGFmdGVyIHI1MzExMApJbmRleDogV2ViQ29yZS9jc3MvQ1NTU2VsZWN0b3IuaAo9PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09Ci0tLSBXZWJDb3JlL2Nzcy9DU1NTZWxlY3Rvci5oCShyZXZpc2lvbiA1MzE0MCkKKysr IFdlYkNvcmUvY3NzL0NTU1NlbGVjdG9yLmgJKHdvcmtpbmcgY29weSkKQEAgLTU4LDcgKzU4LDcg QEAgbmFtZXNwYWNlIFdlYkNvcmUgewogICAgICAgICAgICAgaWYgKG1faGFzUmFyZURhdGEpCiAg ICAgICAgICAgICAgICAgZGVsZXRlIG1fZGF0YS5tX3JhcmVEYXRhOwogICAgICAgICAgICAgZWxz ZQotICAgICAgICAgICAgICAgIGRlbGV0ZSBtX2RhdGEubV90YWdIaXN0b3J5OworICAgICAgICAg ICAgICAgIGZhc3REZWxldGUobV9kYXRhLm1fdGFnSGlzdG9yeSk7CiAgICAgICAgIH0KIAogICAg ICAgICAvKioKSW5kZXg6IFdlYkNvcmUvY3NzL0NTU1NlbGVjdG9yTGlzdC5jcHAKPT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PQotLS0gV2ViQ29yZS9jc3MvQ1NTU2VsZWN0b3JMaXN0LmNwcAkocmV2aXNpb24gNTMxNDApCisr KyBXZWJDb3JlL2Nzcy9DU1NTZWxlY3Rvckxpc3QuY3BwCSh3b3JraW5nIGNvcHkpCkBAIC03Niw3 ICs3Niw3IEBAIHZvaWQgQ1NTU2VsZWN0b3JMaXN0OjpkZWxldGVTZWxlY3RvcnMoKQogICAgIENT U1NlbGVjdG9yKiBzID0gbV9zZWxlY3RvckFycmF5OwogICAgIGJvb2wgZG9uZSA9IHMtPmlzTGFz dEluU2VsZWN0b3JMaXN0KCk7CiAgICAgaWYgKGRvbmUpCi0gICAgICAgIGRlbGV0ZSBzOworICAg ICAgICBmYXN0RGVsZXRlKHMpOwogICAgIGVsc2UgewogICAgICAgICB3aGlsZSAoMSkgewogICAg ICAgICAgICAgcy0+fkNTU1NlbGVjdG9yKCk7Cg== 46376 2010-01-12 07:55:09 -0800 2010-06-11 10:40:45 -0700 Patch for comment #23 (retry) 22834.patch text/plain 1766 bnason SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiA1MzE0MCkKKysrIFdlYkNvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBA IC0xLDMgKzEsMTkgQEAKKzIwMTAtMDEtMTIgIEJlbmJ1Y2sgTmFzb24gIDxibmFzb25AbmV0Zmxp eC5jb20+CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAg aHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTIyODM0CisgICAgICAgIE1p c21hdGNoZWQgbWVtb3J5IGZyZWUgaW4gQ1NTU2VsZWN0b3JMaXN0CisKKyAgICAgICAgRml4IG1p c21hdGNoZWQgZmFzdE1hbGxvYygpL2RlbGV0ZSBpbiBDU1NTZWxlY3Rvckxpc3Q6OmRlbGV0ZVNl bGVjdG9ycygpCisgICAgICAgIGFuZCBDU1NTZWxlY3Rvcjo6fkNTU1NlbGVjdG9yKCkgd2hlbiBG QVNUX0ZBU1RfTUFMTE9DX01BVENIX1ZBTElEQVRJT049MQorICAgICAgICBpcyBzZXQuCisKKyAg ICAgICAgKiBjc3MvQ1NTU2VsZWN0b3IuaDoKKyAgICAgICAgKFdlYkNvcmU6OkNTU1NlbGVjdG9y Ojp+Q1NTU2VsZWN0b3IpOgorICAgICAgICAqIGNzcy9DU1NTZWxlY3Rvckxpc3QuY3BwOgorICAg ICAgICAoV2ViQ29yZTo6Q1NTU2VsZWN0b3JMaXN0OjpkZWxldGVTZWxlY3RvcnMpOgorCiAyMDEw LTAxLTEyICBBZGFtIFJvYmVuICA8YXJvYmVuQGFwcGxlLmNvbT4KIAogICAgICAgICBXaW5kb3dz IGFjY2VsZXJhdGVkIGNvbXBvc2l0aW5nIGJ1aWxkIGZpeCBhZnRlciByNTMxMTAKSW5kZXg6IFdl YkNvcmUvY3NzL0NTU1NlbGVjdG9yLmgKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gV2ViQ29yZS9jc3MvQ1NTU2Vs ZWN0b3IuaAkocmV2aXNpb24gNTMxNDApCisrKyBXZWJDb3JlL2Nzcy9DU1NTZWxlY3Rvci5oCSh3 b3JraW5nIGNvcHkpCkBAIC01OCw3ICs1OCw3IEBAIG5hbWVzcGFjZSBXZWJDb3JlIHsKICAgICAg ICAgICAgIGlmIChtX2hhc1JhcmVEYXRhKQogICAgICAgICAgICAgICAgIGRlbGV0ZSBtX2RhdGEu bV9yYXJlRGF0YTsKICAgICAgICAgICAgIGVsc2UKLSAgICAgICAgICAgICAgICBkZWxldGUgbV9k YXRhLm1fdGFnSGlzdG9yeTsKKyAgICAgICAgICAgICAgICBmYXN0RGVsZXRlKG1fZGF0YS5tX3Rh Z0hpc3RvcnkpOwogICAgICAgICB9CiAKICAgICAgICAgLyoqCkluZGV4OiBXZWJDb3JlL2Nzcy9D U1NTZWxlY3Rvckxpc3QuY3BwCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvY3NzL0NTU1NlbGVjdG9y TGlzdC5jcHAJKHJldmlzaW9uIDUzMTQwKQorKysgV2ViQ29yZS9jc3MvQ1NTU2VsZWN0b3JMaXN0 LmNwcAkod29ya2luZyBjb3B5KQpAQCAtNzYsNyArNzYsNyBAQCB2b2lkIENTU1NlbGVjdG9yTGlz dDo6ZGVsZXRlU2VsZWN0b3JzKCkKICAgICBDU1NTZWxlY3RvciogcyA9IG1fc2VsZWN0b3JBcnJh eTsKICAgICBib29sIGRvbmUgPSBzLT5pc0xhc3RJblNlbGVjdG9yTGlzdCgpOwogICAgIGlmIChk b25lKQotICAgICAgICBkZWxldGUgczsKKyAgICAgICAgZmFzdERlbGV0ZShzKTsKICAgICBlbHNl IHsKICAgICAgICAgd2hpbGUgKDEpIHsKICAgICAgICAgICAgIHMtPn5DU1NTZWxlY3RvcigpOwo=