58027 2011-04-07 03:59:42 -0700 Avoid leaking document when leaving google.com due to geolocation permission request 2011-10-19 17:27:45 -0700 1 1 1 Unclassified WebKit WebCore Misc. 528+ (Nightly build) PC All RESOLVED FIXED P2 Normal --- 1 kenneth satish jknotten kling satish steveblock tonikitoo webkit.review.bot oldest_to_newest 381226 0 kenneth 2011-04-07 03:59:42 -0700 SSIA 381229 1 88606 kenneth 2011-04-07 04:08:49 -0700 Created attachment 88606 Patch 381232 2 88606 jknotten 2011-04-07 04:24:34 -0700 Comment on attachment 88606 Patch This seems correct to me. If we are removing the notifiers from the oneShots and watchers list, it also makes sense to remove them from the other lists too. 382144 3 88606 steveblock 2011-04-08 04:35:28 -0700 Comment on attachment 88606 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=88606&action=review > Source/WebCore/ChangeLog:11 > + In fatalErrorOccurred (which is called on cancellation), the notifier What do you mean by 'cancellation'? Can you elaborate in the bug exactly what the problem is? Presumably it's not unique to google.com? > Source/WebCore/ChangeLog:18 > + Can you add a test for this? If not, you should explain here why not. > Source/WebCore/page/Geolocation.cpp:342 > + m_requestsAwaitingCachedPosition.remove(notifier); I think that the only way for a notifier to be in this list when it encounters a fatal error is when Geolocation::reset() is called. Is this what you're protecting against? 382158 4 kenneth 2011-04-08 05:08:13 -0700 (In reply to comment #3) > (From update of attachment 88606 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=88606&action=review > > > Source/WebCore/ChangeLog:11 > > + In fatalErrorOccurred (which is called on cancellation), the notifier > > What do you mean by 'cancellation'? Can you elaborate in the bug exactly what the problem is? Presumably it's not unique to google.com? Reloading the page or loading another one, leaks document() probably due to event holding a ref to document (presumable TargetEvent). It probably happens on other pages as well (using watchPosition?), but I didn't manage to trigger the leak with my own simple tests, and debugging what exactly is happening on google.com is pretty hard due to the obscured javascript code. > > > Source/WebCore/ChangeLog:18 > > + > > Can you add a test for this? If not, you should explain here why not. The test can only work for ports using the PREEMPT_... build flag, and I am not sure now to actually test this, so any help is appreciated. Testing it is hard due to the reloading, plus the fact that I cannot AFAIK count the document references from LayoutTestController. > > Source/WebCore/page/Geolocation.cpp:342 > > + m_requestsAwaitingCachedPosition.remove(notifier); > > I think that the only way for a notifier to be in this list when it encounters a fatal error is when Geolocation::reset() is called. Is this what you're protecting against? I added this one for completions sake. It can be left out. 382193 5 88606 steveblock 2011-04-08 06:19:15 -0700 Comment on attachment 88606 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=88606&action=review >>> Source/WebCore/ChangeLog:11 >>> + In fatalErrorOccurred (which is called on cancellation), the notifier >> >> What do you mean by 'cancellation'? Can you elaborate in the bug exactly what the problem is? Presumably it's not unique to google.com? > > Reloading the page or loading another one, leaks document() probably due to event holding a ref to document (presumable TargetEvent). > > It probably happens on other pages as well (using watchPosition?), but I didn't manage to trigger the leak with my own simple tests, and debugging what exactly is happening on google.com is pretty hard due to the obscured javascript code. Do you see this only on Chromium? I think it would be good to understand a little more about the problem before we make the fix. The Geolocation callbacks don't ref their ScriptExecutionContext, so I don't see how removing this notifier from this list will avoid a leak. Presumably the Geolocation object is leaked too? 382766 6 kenneth 2011-04-09 04:20:31 -0700 > Do you see this only on Chromium? I think it would be good to understand a little more about the problem before we make the fix. The Geolocation callbacks don't ref their ScriptExecutionContext, so I don't see how removing this notifier from this list will avoid a leak. Presumably the Geolocation object is leaked too? This happens on our Qt WebKit2 port (a branch on gitorious). It was pretty hard to debug because Document is ref'ed for each event, but adding this plugged the leak. 408360 7 88606 eric 2011-05-23 15:09:20 -0700 Comment on attachment 88606 Patch We need to find some way to test this. Manual tests are fine. 409563 8 88606 jknotten 2011-05-25 03:18:22 -0700 Comment on attachment 88606 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=88606&action=review A test to count the number of notifiers left in the lists after page reload should be possible. However, exposing this count seems only useful for this particular test case. Is there a way to add this test without complicating the interfaces of the Geolocation, GeolocationController, GeolocationClient and embedder-specific client classes? >>> Source/WebCore/page/Geolocation.cpp:342 >>> + m_requestsAwaitingCachedPosition.remove(notifier); >> >> I think that the only way for a notifier to be in this list when it encounters a fatal error is when Geolocation::reset() is called. Is this what you're protecting against? > > I added this one for completions sake. It can be left out. Steve: Yes, we can get a build up of notifiers in this case, since the main Frame (and therefore Geolocation instance) is reused across page reloads. 409568 9 jknotten 2011-05-25 03:33:45 -0700 fast/dom/Geolocation/page-reload-cancel-permission-requests.html is a related test that tests that geolocation permission requests are properly cancelled at the GeolocationClient on page reload. (In reply to comment #8) > (From update of attachment 88606 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=88606&action=review > > A test to count the number of notifiers left in the lists after page reload should be possible. However, exposing this count seems only useful for this particular test case. Is there a way to add this test without complicating the interfaces of the Geolocation, GeolocationController, GeolocationClient and embedder-specific client classes? > > >>> Source/WebCore/page/Geolocation.cpp:342 > >>> + m_requestsAwaitingCachedPosition.remove(notifier); > >> > >> I think that the only way for a notifier to be in this list when it encounters a fatal error is when Geolocation::reset() is called. Is this what you're protecting against? > > > > I added this one for completions sake. It can be left out. > > Steve: Yes, we can get a build up of notifiers in this case, since the main Frame (and therefore Geolocation instance) is reused across page reloads. 483991 10 satish 2011-10-14 08:28:38 -0700 I've noticed the same issue. I've tested only with Chromium so not sure if it happens with other browsers as well. FWIW, I found adding the following to the Geolocation::reset() method works as well. #if USE(PREEMPT_GEOLOCATION_PERMISSION) m_pendingForPermissionNotifiers.remove(notifier); #endif 483992 11 satish 2011-10-14 08:31:44 -0700 I can also confirm that the leaking of Document was due to these pending GeoNotifier objects. In chromium these were creating a persistent v8 handle which weren't released properly (as these were still alive when the page closed) caused the whole DOM to be leaked when the page closed. 484003 12 satish 2011-10-14 08:46:12 -0700 Sorry, I should've mentioned that I used this code in the reset() method: #if USE(PREEMPT_GEOLOCATION_PERMISSION) m_pendingForPermissionNotifiers.clear(); #endif 484015 13 jknotten 2011-10-14 09:26:03 -0700 I think we should also be removing the watch from m_pendingForPermissionNotifiers in Geolocation::clearWatch to avoid potential build up of watches by repeated calls to watchPosition and clearWatch. 484559 14 111130 satish 2011-10-15 02:11:51 -0700 Created attachment 111130 Patch 484573 15 111130 kenneth 2011-10-15 05:52:12 -0700 Comment on attachment 111130 Patch Nice, but can you implement John's suggestion as well? It could be a follow up. 484600 16 111130 jknotten 2011-10-15 10:25:53 -0700 Comment on attachment 111130 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111130&action=review > Source/WebCore/page/Geolocation.cpp:344 > +#if USE(PREEMPT_GEOLOCATION_PERMISSION) The change to Geolocation::reset() is enough - m_pendingForPermissionNotifiers is cleared in Geolocation::setIsAllowed. All other calls to setFatalError are called after permission has been determined, so the notifier won't be in this list. It would be good to move these lines to Geolocation::clearWatch(), though, to avoid a potential build up of watches in this set. 484755 17 111130 satish 2011-10-16 02:18:02 -0700 Comment on attachment 111130 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111130&action=review >> Source/WebCore/page/Geolocation.cpp:344 >> +#if USE(PREEMPT_GEOLOCATION_PERMISSION) > > The change to Geolocation::reset() is enough - m_pendingForPermissionNotifiers is cleared in Geolocation::setIsAllowed. All other calls to setFatalError are called after permission has been determined, so the notifier won't be in this list. > > It would be good to move these lines to Geolocation::clearWatch(), though, to avoid a potential build up of watches in this set. Will do. 484757 18 111178 satish 2011-10-16 02:24:03 -0700 Created attachment 111178 Patch 484758 19 111178 kenneth 2011-10-16 02:33:45 -0700 Comment on attachment 111178 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111178&action=review > Source/WebCore/page/Geolocation.cpp:178 > +Geolocation::GeoNotifier* Geolocation::Watchers::get(int id) WebKit style would be to call this method for "find" 484759 20 111178 kenneth 2011-10-16 02:35:58 -0700 Comment on attachment 111178 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111178&action=review > Source/WebCore/page/Geolocation.cpp:436 > + GeoNotifier* notifier = m_watchers.get(watchId); > + if (notifier) > + m_pendingForPermissionNotifiers.remove(notifier); You could change this into: if (GeoNotifier* notifier = m_watchers.find(watchId)) m_pendingForPermissionNotifiers.remove(notifier); 484760 21 kenneth 2011-10-16 02:40:20 -0700 > > The change to Geolocation::reset() is enough - m_pendingForPermissionNotifiers is cleared in Geolocation::setIsAllowed. I think it would be better do document such behavior using ASSERTs if possible. 484817 22 111178 satish 2011-10-16 14:13:42 -0700 Comment on attachment 111178 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111178&action=review >> Source/WebCore/page/Geolocation.cpp:178 >> +Geolocation::GeoNotifier* Geolocation::Watchers::get(int id) > > WebKit style would be to call this method for "find" Will do >> Source/WebCore/page/Geolocation.cpp:436 >> + m_pendingForPermissionNotifiers.remove(notifier); > > You could change this into: > > if (GeoNotifier* notifier = m_watchers.find(watchId)) > m_pendingForPermissionNotifiers.remove(notifier); Will do 484818 23 111186 satish 2011-10-16 14:15:26 -0700 Created attachment 111186 Patch 484992 24 111186 kenneth 2011-10-17 01:39:11 -0700 Comment on attachment 111186 Patch John are you OK with this? 484993 25 jknotten 2011-10-17 01:43:53 -0700 (In reply to comment #24) > (From update of attachment 111186 [details]) > John are you OK with this? Looks good to me. thanks for reviewing! 485000 26 111186 webkit.review.bot 2011-10-17 02:01:29 -0700 Comment on attachment 111186 Patch Clearing flags on attachment: 111186 Committed r97594: <http://trac.webkit.org/changeset/97594> 485001 27 webkit.review.bot 2011-10-17 02:01:38 -0700 All reviewed patches have been landed. Closing bug. 485176 28 111186 steveblock 2011-10-17 08:32:58 -0700 Comment on attachment 111186 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111186&action=review > Source/WebCore/page/Geolocation.h:116 > + GeoNotifier* find(int id); Do we need this new method? All we do with the GeoNotifier returned by this method is pass it to remove(GeoNotifier*), but Watchers already offers a remove(int id) method, which achieves the same thing. 485178 29 111186 satish 2011-10-17 08:38:46 -0700 Comment on attachment 111186 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111186&action=review >> Source/WebCore/page/Geolocation.h:116 >> + GeoNotifier* find(int id); > > Do we need this new method? All we do with the GeoNotifier returned by this method is pass it to remove(GeoNotifier*), but Watchers already offers a remove(int id) method, which achieves the same thing. You're right, I don't know how I missed that :/ Will upload a patch to remove this method and use remove(int id) instead. 485182 30 111186 satish 2011-10-17 08:45:09 -0700 Comment on attachment 111186 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=111186&action=review >>> Source/WebCore/page/Geolocation.h:116 >>> + GeoNotifier* find(int id); >> >> Do we need this new method? All we do with the GeoNotifier returned by this method is pass it to remove(GeoNotifier*), but Watchers already offers a remove(int id) method, which achieves the same thing. > > You're right, I don't know how I missed that :/ Will upload a patch to remove this method and use remove(int id) instead. m_pendingForPermissionNotifiers is not part of m_watchers. So calling Watchers::remove(int id) can't remove this pending item. So leaving it as is. 485199 31 steveblock 2011-10-17 09:14:26 -0700 >>>> Source/WebCore/page/Geolocation.cpp:342 >>>> + m_requestsAwaitingCachedPosition.remove(notifier); >>> >>> I think that the only way for a notifier to be in this list when it encounters a fatal error is when Geolocation::reset() is called. Is this what you're protecting against? >> I added this one for completions sake. It can be left out. > Steve: Yes, we can get a build up of notifiers in this case, since the main Frame (and therefore Geolocation instance) is reused across page reloads. John, are you still concerned about this? How exactly do we handle cancelling regular requests in this case? It looks like the only time we do any cancelling of requests is when reset() or disconectFrame() is called. Is either of these called in this case? If you think there's a problem, let's open a new bug, preferably with a test case. 485248 32 jknotten 2011-10-17 10:17:54 -0700 Steve, are you asking about my earlier proposed change to Geolocation::fatalErrorOccurred? If so, I think it's not necessary to make any changes there now as they will be cleared up in ::reset() with Satish's change. (In reply to comment #31) > >>>> Source/WebCore/page/Geolocation.cpp:342 > >>>> + m_requestsAwaitingCachedPosition.remove(notifier); > >>> > >>> I think that the only way for a notifier to be in this list when it encounters a fatal error is when Geolocation::reset() is called. Is this what you're protecting against? > >> I added this one for completions sake. It can be left out. > > Steve: Yes, we can get a build up of notifiers in this case, since the main Frame (and therefore Geolocation instance) is reused across page reloads. > > John, are you still concerned about this? How exactly do we handle cancelling regular requests in this case? It looks like the only time we do any cancelling of requests is when reset() or disconectFrame() is called. Is either of these called in this case? If you think there's a problem, let's open a new bug, preferably with a test case. 486867 33 steveblock 2011-10-19 10:03:18 -0700 Yes, I was talking about your comment about clearing m_requestsAwaitingCachedPosition in fatalErrorOccurred() in the first patch set uploaded to this bug. Satish's patch doesn't touch m_requestsAwaitingCachedPosition, right? 487167 34 jknotten 2011-10-19 17:27:45 -0700 Steve, I'm not sure if there is a problem there now. I've opened https://bugs.webkit.org/show_bug.cgi?id=70461 to track an investigation of the question. 88606 2011-04-07 04:08:49 -0700 2011-10-15 02:11:39 -0700 Patch 0001-Avoid-leaking-document-when-leaving-google.com-due-t.patch text/plain 2497 kenneth RnJvbSBmYzFiNDdhNDFjMDgzZWM1MzYyZmRhMGRjODA0Njk2NjBiZGE2N2M5IE1vbiBTZXAgMTcg MDA6MDA6MDAgMjAwMQpGcm9tOiBLZW5uZXRoIFJvaGRlIENocmlzdGlhbnNlbiA8a2VubmV0aEB3 ZWJraXQub3JnPgpEYXRlOiBUaHUsIDcgQXByIDIwMTEgMTI6NTQ6MjIgKzAyMDAKU3ViamVjdDog W1BBVENIXSBBdm9pZCBsZWFraW5nIGRvY3VtZW50IHdoZW4gbGVhdmluZyBnb29nbGUuY29tIGR1 ZSB0byBnZW9sb2NhdGlvbgogcGVybWlzc2lvbiByZXF1ZXN0CiBodHRwczovL2J1Z3Mud2Via2l0 Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9NTgwMjcKClJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgoK RG9uZSBpbiBjb29wZXJhdGlvbiB3aXRoIEpvaG4gS25vdHRlbmJlbHQgPGprbm90dGVuQGNocm9t aXVtLm9yZz4KCkluIGZhdGFsRXJyb3JPY2N1cnJlZCAod2hpY2ggaXMgY2FsbGVkIG9uIGNhbmNl bGxhdGlvbiksIHRoZSBub3RpZmllcgp3YXMgbmVpdGhlciByZW1vdmVkIGZyb20gbV9wZW5kaW5n Rm9yUGVybWlzc2lvbk5vdGlmaWVycyBub3IKbV9yZXF1ZXN0c0F3YWl0aW5nQ2FjaGVkUG9zaXRp b24uIFRoZSBmb3JtZXIgaXMgdXNlZCBvbmx5IHdoZW4KUFJFRU1QVF9HRU9MT0NBVElPTl9QRVJN SVNTSU9OIGlzIHVzZWQuCgoqIHBhZ2UvR2VvbG9jYXRpb24uY3BwOgooV2ViQ29yZTo6R2VvbG9j YXRpb246OmZhdGFsRXJyb3JPY2N1cnJlZCk6Ci0tLQogU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9n ICAgICAgICAgICAgfCAgIDE4ICsrKysrKysrKysrKysrKysrKwogU291cmNlL1dlYkNvcmUvcGFn ZS9HZW9sb2NhdGlvbi5jcHAgfCAgICA0ICsrKysKIDIgZmlsZXMgY2hhbmdlZCwgMjIgaW5zZXJ0 aW9ucygrKSwgMCBkZWxldGlvbnMoLSkKCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9DaGFu Z2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggZWViNmJlOC4uZmMyOWMwMCAx MDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9XZWJDb3Jl L0NoYW5nZUxvZwpAQCAtMSwzICsxLDIxIEBACisyMDExLTA0LTA3ICBLZW5uZXRoIFJvaGRlIENo cmlzdGlhbnNlbiAgPGtlbm5ldGhAd2Via2l0Lm9yZz4KKworICAgICAgICBSZXZpZXdlZCBieSBO T0JPRFkgKE9PUFMhKS4KKworICAgICAgICBBdm9pZCBsZWFraW5nIGRvY3VtZW50IHdoZW4gbGVh dmluZyBnb29nbGUuY29tIGR1ZSB0byBnZW9sb2NhdGlvbgorICAgICAgICBwZXJtaXNzaW9uIHJl cXVlc3QKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTU4 MDI3CisKKyAgICAgICAgRG9uZSBpbiBjb29wZXJhdGlvbiB3aXRoIEpvaG4gS25vdHRlbmJlbHQg PGprbm90dGVuQGNocm9taXVtLm9yZz4KKworICAgICAgICBJbiBmYXRhbEVycm9yT2NjdXJyZWQg KHdoaWNoIGlzIGNhbGxlZCBvbiBjYW5jZWxsYXRpb24pLCB0aGUgbm90aWZpZXIKKyAgICAgICAg d2FzIG5laXRoZXIgcmVtb3ZlZCBmcm9tIG1fcGVuZGluZ0ZvclBlcm1pc3Npb25Ob3RpZmllcnMg bm9yCisgICAgICAgIG1fcmVxdWVzdHNBd2FpdGluZ0NhY2hlZFBvc2l0aW9uLiBUaGUgZm9ybWVy IGlzIHVzZWQgb25seSB3aGVuCisgICAgICAgIFBSRUVNUFRfR0VPTE9DQVRJT05fUEVSTUlTU0lP TiBpcyB1c2VkLgorCisgICAgICAgICogcGFnZS9HZW9sb2NhdGlvbi5jcHA6CisgICAgICAgIChX ZWJDb3JlOjpHZW9sb2NhdGlvbjo6ZmF0YWxFcnJvck9jY3VycmVkKToKKwogMjAxMS0wNC0wNyAg QWxpY2UgQm94aGFsbCAgPGFib3hoYWxsQGNocm9taXVtLm9yZz4KIAogICAgICAgICBSZXZpZXdl ZCBieSBSeW9zdWtlIE5pd2EuCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9wYWdlL0dlb2xv Y2F0aW9uLmNwcCBiL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9jYXRpb24uY3BwCmluZGV4IDJh NDYzNWYuLjRmMzVlMTggMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9jYXRp b24uY3BwCisrKyBiL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9jYXRpb24uY3BwCkBAIC0zMzYs NiArMzM2LDEwIEBAIHZvaWQgR2VvbG9jYXRpb246OmZhdGFsRXJyb3JPY2N1cnJlZChHZW9sb2Nh dGlvbjo6R2VvTm90aWZpZXIqIG5vdGlmaWVyKQogICAgIC8vIFRoaXMgcmVxdWVzdCBoYXMgZmFp bGVkIGZhdGFsbHkuIFJlbW92ZSBpdCBmcm9tIG91ciBsaXN0cy4KICAgICBtX29uZVNob3RzLnJl bW92ZShub3RpZmllcik7CiAgICAgbV93YXRjaGVycy5yZW1vdmUobm90aWZpZXIpOworI2lmIFVT RShQUkVFTVBUX0dFT0xPQ0FUSU9OX1BFUk1JU1NJT04pCisgICAgbV9wZW5kaW5nRm9yUGVybWlz c2lvbk5vdGlmaWVycy5yZW1vdmUobm90aWZpZXIpOworI2VuZGlmCisgICAgbV9yZXF1ZXN0c0F3 YWl0aW5nQ2FjaGVkUG9zaXRpb24ucmVtb3ZlKG5vdGlmaWVyKTsKIAogICAgIGlmICghaGFzTGlz dGVuZXJzKCkpCiAgICAgICAgIHN0b3BVcGRhdGluZygpOwotLSAKMS43LjEKCg== 111130 2011-10-15 02:11:51 -0700 2011-10-16 02:23:57 -0700 Patch bug-58027-20111015101149.patch text/plain 1619 satish ZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZyBiL1NvdXJjZS9XZWJDb3JlL0No YW5nZUxvZwppbmRleCBjZWY4OTBjYjRmMjQ5OTJiZDFiZGY1OTE0NDEzMTM2N2I2MWEzNDYyLi5h NDY5MjJkYmIxZTVhODFmNzc4YmRjODI2MWZkMWVhNGE1OWNhYjczIDEwMDY0NAotLS0gYS9Tb3Vy Y2UvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCkBAIC0x LDMgKzEsMTUgQEAKKzIwMTEtMTAtMTUgIFNhdGlzaCBTYW1wYXRoICA8c2F0aXNoQGNocm9taXVt Lm9yZz4KKworICAgICAgICBBdm9pZCBsZWFraW5nIGRvY3VtZW50IHdoZW4gbGVhdmluZyBnb29n bGUuY29tIGR1ZSB0byBnZW9sb2NhdGlvbgorICAgICAgICBwZXJtaXNzaW9uIHJlcXVlc3QuCisg ICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD01ODAyNworCisg ICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogcGFnZS9HZW9s b2NhdGlvbi5jcHA6CisgICAgICAgIChXZWJDb3JlOjpHZW9sb2NhdGlvbjo6cmVzZXQpOgorICAg ICAgICAoV2ViQ29yZTo6R2VvbG9jYXRpb246OmZhdGFsRXJyb3JPY2N1cnJlZCk6CisKIDIwMTEt MTAtMTQgIEFkYW0gQmFydGggIDxhYmFydGhAd2Via2l0Lm9yZz4KIAogICAgICAgICBVcGRhdGUg ZXhwZWN0ZWQgcmVzdWx0cy4KZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9j YXRpb24uY3BwIGIvU291cmNlL1dlYkNvcmUvcGFnZS9HZW9sb2NhdGlvbi5jcHAKaW5kZXggMTlm OWFjN2IwMGU4MDU1NTVhM2JhZjI4MmZiODUyYThlYjc0NjExZC4uNjVmYWQ4NjJjNzRkYmZiYjM2 ZjBhMWY0NThjYjc0M2QzMWEzNzNjOSAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvcGFnZS9H ZW9sb2NhdGlvbi5jcHAKKysrIGIvU291cmNlL1dlYkNvcmUvcGFnZS9HZW9sb2NhdGlvbi5jcHAK QEAgLTI1Myw2ICsyNTMsOSBAQCB2b2lkIEdlb2xvY2F0aW9uOjpyZXNldCgpCiAgICAgbV9hbGxv d0dlb2xvY2F0aW9uID0gVW5rbm93bjsKICAgICBjYW5jZWxBbGxSZXF1ZXN0cygpOwogICAgIHN0 b3BVcGRhdGluZygpOworI2lmIFVTRShQUkVFTVBUX0dFT0xPQ0FUSU9OX1BFUk1JU1NJT04pCisg ICAgbV9wZW5kaW5nRm9yUGVybWlzc2lvbk5vdGlmaWVycy5jbGVhcigpOworI2VuZGlmCiB9CiAK IHZvaWQgR2VvbG9jYXRpb246OmRpc2Nvbm5lY3RGcmFtZSgpCkBAIC0zMzgsNiArMzQxLDkgQEAg dm9pZCBHZW9sb2NhdGlvbjo6ZmF0YWxFcnJvck9jY3VycmVkKEdlb2xvY2F0aW9uOjpHZW9Ob3Rp Zmllciogbm90aWZpZXIpCiAgICAgLy8gVGhpcyByZXF1ZXN0IGhhcyBmYWlsZWQgZmF0YWxseS4g UmVtb3ZlIGl0IGZyb20gb3VyIGxpc3RzLgogICAgIG1fb25lU2hvdHMucmVtb3ZlKG5vdGlmaWVy KTsKICAgICBtX3dhdGNoZXJzLnJlbW92ZShub3RpZmllcik7CisjaWYgVVNFKFBSRUVNUFRfR0VP TE9DQVRJT05fUEVSTUlTU0lPTikKKyAgICBtX3BlbmRpbmdGb3JQZXJtaXNzaW9uTm90aWZpZXJz LnJlbW92ZShub3RpZmllcik7CisjZW5kaWYKIAogICAgIGlmICghaGFzTGlzdGVuZXJzKCkpCiAg ICAgICAgIHN0b3BVcGRhdGluZygpOwo= 111178 2011-10-16 02:24:03 -0700 2011-10-16 14:15:19 -0700 Patch bug-58027-20111016102402.patch text/plain 2794 satish ZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZyBiL1NvdXJjZS9XZWJDb3JlL0No YW5nZUxvZwppbmRleCBjZWY4OTBjYjRmMjQ5OTJiZDFiZGY1OTE0NDEzMTM2N2I2MWEzNDYyLi43 NmRhNjEyN2VmNTdhMjg3NGY5NjA0NzM3ZGMzNWZjN2FjZDE5YWM0IDEwMDY0NAotLS0gYS9Tb3Vy Y2UvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCkBAIC0x LDMgKzEsMTggQEAKKzIwMTEtMTAtMTUgIFNhdGlzaCBTYW1wYXRoICA8c2F0aXNoQGNocm9taXVt Lm9yZz4KKworICAgICAgICBBdm9pZCBsZWFraW5nIGRvY3VtZW50IHdoZW4gbGVhdmluZyBnb29n bGUuY29tIGR1ZSB0byBnZW9sb2NhdGlvbgorICAgICAgICBwZXJtaXNzaW9uIHJlcXVlc3QuCisg ICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD01ODAyNworCisg ICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogcGFnZS9HZW9s b2NhdGlvbi5jcHA6CisgICAgICAgIChXZWJDb3JlOjpHZW9sb2NhdGlvbjo6V2F0Y2hlcnM6Omdl dCk6IGFkZGVkCisgICAgICAgIChXZWJDb3JlOjpHZW9sb2NhdGlvbjo6cmVzZXQpOiBjbGVhcmVk IGFsbCBwZW5kaW5nIG5vdGlmaWVycy4KKyAgICAgICAgKFdlYkNvcmU6Okdlb2xvY2F0aW9uOjpj bGVhcldhdGNoKTogcmVtb3ZlZCBmcm9tIHBlbmRpbmcgbm90aWZpZXIgc2V0LgorICAgICAgICAq IHBhZ2UvR2VvbG9jYXRpb24uaDoKKyAgICAgICAgKFdlYkNvcmU6Okdlb2xvY2F0aW9uOjpXYXRj aGVyczo6Z2V0KTogYWRkZWQKKwogMjAxMS0xMC0xNCAgQWRhbSBCYXJ0aCAgPGFiYXJ0aEB3ZWJr aXQub3JnPgogCiAgICAgICAgIFVwZGF0ZSBleHBlY3RlZCByZXN1bHRzLgpkaWZmIC0tZ2l0IGEv U291cmNlL1dlYkNvcmUvcGFnZS9HZW9sb2NhdGlvbi5jcHAgYi9Tb3VyY2UvV2ViQ29yZS9wYWdl L0dlb2xvY2F0aW9uLmNwcAppbmRleCAxOWY5YWM3YjAwZTgwNTU1NWEzYmFmMjgyZmI4NTJhOGVi NzQ2MTFkLi5lMTdiMDJjNmEwYzE5MzFmYjMyNDdkMGNjZmRjMzNhZjExN2FiNjQ2IDEwMDY0NAot LS0gYS9Tb3VyY2UvV2ViQ29yZS9wYWdlL0dlb2xvY2F0aW9uLmNwcAorKysgYi9Tb3VyY2UvV2Vi Q29yZS9wYWdlL0dlb2xvY2F0aW9uLmNwcApAQCAtMTc1LDYgKzE3NSwxNSBAQCB2b2lkIEdlb2xv Y2F0aW9uOjpXYXRjaGVyczo6c2V0KGludCBpZCwgUGFzc1JlZlB0cjxHZW9Ob3RpZmllcj4gcHJw Tm90aWZpZXIpCiAgICAgbV9ub3RpZmllclRvSWRNYXAuc2V0KG5vdGlmaWVyLnJlbGVhc2UoKSwg aWQpOwogfQogCitHZW9sb2NhdGlvbjo6R2VvTm90aWZpZXIqIEdlb2xvY2F0aW9uOjpXYXRjaGVy czo6Z2V0KGludCBpZCkKK3sKKyAgICBBU1NFUlQoaWQgPiAwKTsKKyAgICBJZFRvTm90aWZpZXJN YXA6Oml0ZXJhdG9yIGl0ZXIgPSBtX2lkVG9Ob3RpZmllck1hcC5maW5kKGlkKTsKKyAgICBpZiAo aXRlciA9PSBtX2lkVG9Ob3RpZmllck1hcC5lbmQoKSkKKyAgICAgICAgcmV0dXJuIDA7CisgICAg cmV0dXJuIGl0ZXItPnNlY29uZC5nZXQoKTsKK30KKwogdm9pZCBHZW9sb2NhdGlvbjo6V2F0Y2hl cnM6OnJlbW92ZShpbnQgaWQpCiB7CiAgICAgQVNTRVJUKGlkID4gMCk7CkBAIC0yNTMsNiArMjYy LDkgQEAgdm9pZCBHZW9sb2NhdGlvbjo6cmVzZXQoKQogICAgIG1fYWxsb3dHZW9sb2NhdGlvbiA9 IFVua25vd247CiAgICAgY2FuY2VsQWxsUmVxdWVzdHMoKTsKICAgICBzdG9wVXBkYXRpbmcoKTsK KyNpZiBVU0UoUFJFRU1QVF9HRU9MT0NBVElPTl9QRVJNSVNTSU9OKQorICAgIG1fcGVuZGluZ0Zv clBlcm1pc3Npb25Ob3RpZmllcnMuY2xlYXIoKTsKKyNlbmRpZgogfQogCiB2b2lkIEdlb2xvY2F0 aW9uOjpkaXNjb25uZWN0RnJhbWUoKQpAQCAtNDE4LDYgKzQzMCwxMSBAQCB2b2lkIEdlb2xvY2F0 aW9uOjpjbGVhcldhdGNoKGludCB3YXRjaElkKQogICAgIGlmICh3YXRjaElkIDwgZmlyc3RBdmFp bGFibGVXYXRjaElkKQogICAgICAgICByZXR1cm47CiAKKyNpZiBVU0UoUFJFRU1QVF9HRU9MT0NB VElPTl9QRVJNSVNTSU9OKQorICAgIEdlb05vdGlmaWVyKiBub3RpZmllciA9IG1fd2F0Y2hlcnMu Z2V0KHdhdGNoSWQpOworICAgIGlmIChub3RpZmllcikKKyAgICAgICAgbV9wZW5kaW5nRm9yUGVy bWlzc2lvbk5vdGlmaWVycy5yZW1vdmUobm90aWZpZXIpOworI2VuZGlmCiAgICAgbV93YXRjaGVy cy5yZW1vdmUod2F0Y2hJZCk7CiAgICAgCiAgICAgaWYgKCFoYXNMaXN0ZW5lcnMoKSkKZGlmZiAt LWdpdCBhL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9jYXRpb24uaCBiL1NvdXJjZS9XZWJDb3Jl L3BhZ2UvR2VvbG9jYXRpb24uaAppbmRleCAwNGI5MGRiZjQxNWMwNTUyMjg4ZWU3YTQ2NjIyZWJi MmNjNzY1M2Y2Li4wZWE1YjE1MzEzZjg0YmRkZDE0MjRmMDUxODhlMGM4ZjNlZTI4ODIwIDEwMDY0 NAotLS0gYS9Tb3VyY2UvV2ViQ29yZS9wYWdlL0dlb2xvY2F0aW9uLmgKKysrIGIvU291cmNlL1dl YkNvcmUvcGFnZS9HZW9sb2NhdGlvbi5oCkBAIC0xMTMsNiArMTEzLDcgQEAgcHJpdmF0ZToKICAg ICBjbGFzcyBXYXRjaGVycyB7CiAgICAgcHVibGljOgogICAgICAgICB2b2lkIHNldChpbnQgaWQs IFBhc3NSZWZQdHI8R2VvTm90aWZpZXI+KTsKKyAgICAgICAgR2VvTm90aWZpZXIqIGdldChpbnQg aWQpOwogICAgICAgICB2b2lkIHJlbW92ZShpbnQgaWQpOwogICAgICAgICB2b2lkIHJlbW92ZShH ZW9Ob3RpZmllciopOwogICAgICAgICBib29sIGNvbnRhaW5zKEdlb05vdGlmaWVyKikgY29uc3Q7 Cg== 111186 2011-10-16 14:15:26 -0700 2011-10-17 08:45:08 -0700 Patch bug-58027-20111016221524.patch text/plain 2808 satish ZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZyBiL1NvdXJjZS9XZWJDb3JlL0No YW5nZUxvZwppbmRleCA3Yjc3ODJlZDJlZmM4ZGY2YWFiZWZmMTc3MDQyN2QyZTVhNDU1ZjBiLi41 NjBlOWViMTMzYjIyYmQxOTFmYTE3YmNhODBkOTUwZmFjZjkwYjlkIDEwMDY0NAotLS0gYS9Tb3Vy Y2UvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCkBAIC0x LDMgKzEsMTggQEAKKzIwMTEtMTAtMTUgIFNhdGlzaCBTYW1wYXRoICA8c2F0aXNoQGNocm9taXVt Lm9yZz4KKworICAgICAgICBBdm9pZCBsZWFraW5nIGRvY3VtZW50IHdoZW4gbGVhdmluZyBnb29n bGUuY29tIGR1ZSB0byBnZW9sb2NhdGlvbgorICAgICAgICBwZXJtaXNzaW9uIHJlcXVlc3QuCisg ICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD01ODAyNworCisg ICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogcGFnZS9HZW9s b2NhdGlvbi5jcHA6CisgICAgICAgIChXZWJDb3JlOjpHZW9sb2NhdGlvbjo6V2F0Y2hlcnM6OmZp bmQpOiBhZGRlZAorICAgICAgICAoV2ViQ29yZTo6R2VvbG9jYXRpb246OnJlc2V0KTogY2xlYXJl ZCBhbGwgcGVuZGluZyBub3RpZmllcnMuCisgICAgICAgIChXZWJDb3JlOjpHZW9sb2NhdGlvbjo6 Y2xlYXJXYXRjaCk6IHJlbW92ZWQgZnJvbSBwZW5kaW5nIG5vdGlmaWVyIHNldC4KKyAgICAgICAg KiBwYWdlL0dlb2xvY2F0aW9uLmg6CisgICAgICAgIChXZWJDb3JlOjpHZW9sb2NhdGlvbjo6V2F0 Y2hlcnM6OmZpbmQpOiBhZGRlZAorCiAyMDExLTEwLTE2ICBEbWl0cnkgTG9tb3YgIDxkc2xvbW92 QGdvb2dsZS5jb20+CiAKICAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcu Y2dpP2lkPTcwMTg2CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9wYWdlL0dlb2xvY2F0aW9u LmNwcCBiL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9jYXRpb24uY3BwCmluZGV4IDE5ZjlhYzdi MDBlODA1NTU1YTNiYWYyODJmYjg1MmE4ZWI3NDYxMWQuLjFkYWVhNzQ0OTZlNjhjNTBjMTM2OGNj MmQyOTRlMDdlMTk5Nzg4MDQgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9j YXRpb24uY3BwCisrKyBiL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9jYXRpb24uY3BwCkBAIC0x NzUsNiArMTc1LDE1IEBAIHZvaWQgR2VvbG9jYXRpb246OldhdGNoZXJzOjpzZXQoaW50IGlkLCBQ YXNzUmVmUHRyPEdlb05vdGlmaWVyPiBwcnBOb3RpZmllcikKICAgICBtX25vdGlmaWVyVG9JZE1h cC5zZXQobm90aWZpZXIucmVsZWFzZSgpLCBpZCk7CiB9CiAKK0dlb2xvY2F0aW9uOjpHZW9Ob3Rp ZmllciogR2VvbG9jYXRpb246OldhdGNoZXJzOjpmaW5kKGludCBpZCkKK3sKKyAgICBBU1NFUlQo aWQgPiAwKTsKKyAgICBJZFRvTm90aWZpZXJNYXA6Oml0ZXJhdG9yIGl0ZXIgPSBtX2lkVG9Ob3Rp Zmllck1hcC5maW5kKGlkKTsKKyAgICBpZiAoaXRlciA9PSBtX2lkVG9Ob3RpZmllck1hcC5lbmQo KSkKKyAgICAgICAgcmV0dXJuIDA7CisgICAgcmV0dXJuIGl0ZXItPnNlY29uZC5nZXQoKTsKK30K Kwogdm9pZCBHZW9sb2NhdGlvbjo6V2F0Y2hlcnM6OnJlbW92ZShpbnQgaWQpCiB7CiAgICAgQVNT RVJUKGlkID4gMCk7CkBAIC0yNTMsNiArMjYyLDkgQEAgdm9pZCBHZW9sb2NhdGlvbjo6cmVzZXQo KQogICAgIG1fYWxsb3dHZW9sb2NhdGlvbiA9IFVua25vd247CiAgICAgY2FuY2VsQWxsUmVxdWVz dHMoKTsKICAgICBzdG9wVXBkYXRpbmcoKTsKKyNpZiBVU0UoUFJFRU1QVF9HRU9MT0NBVElPTl9Q RVJNSVNTSU9OKQorICAgIG1fcGVuZGluZ0ZvclBlcm1pc3Npb25Ob3RpZmllcnMuY2xlYXIoKTsK KyNlbmRpZgogfQogCiB2b2lkIEdlb2xvY2F0aW9uOjpkaXNjb25uZWN0RnJhbWUoKQpAQCAtNDE4 LDYgKzQzMCwxMCBAQCB2b2lkIEdlb2xvY2F0aW9uOjpjbGVhcldhdGNoKGludCB3YXRjaElkKQog ICAgIGlmICh3YXRjaElkIDwgZmlyc3RBdmFpbGFibGVXYXRjaElkKQogICAgICAgICByZXR1cm47 CiAKKyNpZiBVU0UoUFJFRU1QVF9HRU9MT0NBVElPTl9QRVJNSVNTSU9OKQorICAgIGlmIChHZW9O b3RpZmllciogbm90aWZpZXIgPSBtX3dhdGNoZXJzLmZpbmQod2F0Y2hJZCkpCisgICAgICAgIG1f cGVuZGluZ0ZvclBlcm1pc3Npb25Ob3RpZmllcnMucmVtb3ZlKG5vdGlmaWVyKTsKKyNlbmRpZgog ICAgIG1fd2F0Y2hlcnMucmVtb3ZlKHdhdGNoSWQpOwogICAgIAogICAgIGlmICghaGFzTGlzdGVu ZXJzKCkpCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9wYWdlL0dlb2xvY2F0aW9uLmggYi9T b3VyY2UvV2ViQ29yZS9wYWdlL0dlb2xvY2F0aW9uLmgKaW5kZXggMDRiOTBkYmY0MTVjMDU1MjI4 OGVlN2E0NjYyMmViYjJjYzc2NTNmNi4uYjI3ZDgwYjk4NjEzNjA0ZWZiMmQwY2JhOGE4ZjE0NjI4 OGI3NTlmNiAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvcGFnZS9HZW9sb2NhdGlvbi5oCisr KyBiL1NvdXJjZS9XZWJDb3JlL3BhZ2UvR2VvbG9jYXRpb24uaApAQCAtMTEzLDYgKzExMyw3IEBA IHByaXZhdGU6CiAgICAgY2xhc3MgV2F0Y2hlcnMgewogICAgIHB1YmxpYzoKICAgICAgICAgdm9p ZCBzZXQoaW50IGlkLCBQYXNzUmVmUHRyPEdlb05vdGlmaWVyPik7CisgICAgICAgIEdlb05vdGlm aWVyKiBmaW5kKGludCBpZCk7CiAgICAgICAgIHZvaWQgcmVtb3ZlKGludCBpZCk7CiAgICAgICAg IHZvaWQgcmVtb3ZlKEdlb05vdGlmaWVyKik7CiAgICAgICAgIGJvb2wgY29udGFpbnMoR2VvTm90 aWZpZXIqKSBjb25zdDsK