39475 2010-05-21 00:18:12 -0700 [chromium] if keydown is prevented, don't update the IME and clear the IME state 2010-05-23 18:14:52 -0700 1 1 1 Unclassified WebKit New Bugs 528+ (Nightly build) Other OS X 10.5 RESOLVED FIXED P2 Normal --- 1 tony webkit-unassigned eric hbono ojan oldest_to_newest 228776 0 tony 2010-05-21 00:18:12 -0700 [chromium] if keydown is prevented, don't update the IME and clear the IME state 228777 1 56678 tony 2010-05-21 00:18:40 -0700 Created attachment 56678 Patch 228778 2 tony 2010-05-21 00:19:44 -0700 This is similar to the original patch, but in addition to ignoring the event, we clear out any existing IME state. 229033 3 eric 2010-05-21 12:10:24 -0700 How does this behavior compare to other WebKit ports? Most noteably apple Mac and Apple Win? 229193 4 56678 ojan 2010-05-21 17:52:25 -0700 Comment on attachment 56678 Patch r=me with the added FIXME. Given the total inconsistency of IME preventDefault-handling on Mac browsers and that all Windows browsers seem to properly handle preventDefault in that case, I think it's safe to experiment in the direction of canceling. Also, this is an improvement over the bugs we found from r59735, so I'm OK with committing it and trying it out on the dev channel. WebKit/chromium/src/WebViewImpl.cpp:1193 + if ((command == WebCompositionCommandDiscard) || m_suppressNextKeypressEvent) { Please add a FIXME that we should really try to cancel individual keydowns instead of discarding the entire composition. (In reply to comment #2) > This is similar to the original patch, but in addition to ignoring the event, we clear out any existing IME state. For context, the original patch is http://trac.webkit.org/changeset/59735. This is a better solution than r59735, but it's still not the ideal behavior. Ideally, preventDefault on a keydown would only cancel that key press, not the entire composition. For now, I'm OK with trying this and seeing if it meets Wave's use case. (In reply to comment #3) > How does this behavior compare to other WebKit ports? Most noteably apple Mac and Apple Win? preventDefault of keyDown is a no-op on Apple Mac. As best I can tell, Apple Win actually properly cancels preventDefaulted keydowns. Although, I only tested a few IMEs. There may be Windows IMEs that don't properly preventDefault. It's not clear to me whether the original behavior (preventDefault is a no-op) or the new behavior (discard the composition) is better. Neither one meets the use-case of excluding certain types of characters (e.g. non-numbers), although I'm not really sure there's much use-case for that when you are mid-composition. We should make sure to get actively solicit feedback from the relevant developers (e.g. Wave) and be ready to roll the whole thing back before the next beta/stable release. I'm not sure if there's a use-case for just summarily disallowing any text entry in an editable area. If there is, then the new behavior at least affords that. 229548 5 tony 2010-05-23 18:14:52 -0700 Committed r60054: <http://trac.webkit.org/changeset/60054> 56678 2010-05-21 00:18:40 -0700 2010-05-21 17:52:24 -0700 Patch bug-39475-20100521161838.patch text/plain 2103 tony ZGlmZiAtLWdpdCBhL1dlYktpdC9jaHJvbWl1bS9DaGFuZ2VMb2cgYi9XZWJLaXQvY2hyb21pdW0v Q2hhbmdlTG9nCmluZGV4IDhhNjhjNGU2ZTg0MzQ1MzNhNzVkZjcyNDk1MDVkOGQ3OTI2MGJmM2Eu Ljc1ODdkN2RjZmI2MGUwOTBhNTMyZTU3OWRiNDU4ZDRmMGFiNTEzZWQgMTAwNjQ0Ci0tLSBhL1dl YktpdC9jaHJvbWl1bS9DaGFuZ2VMb2cKKysrIGIvV2ViS2l0L2Nocm9taXVtL0NoYW5nZUxvZwpA QCAtMSwzICsxLDEzIEBACisyMDEwLTA1LTIxICBUb255IENoYW5nICA8dG9ueUBjaHJvbWl1bS5v cmc+CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAgW2No cm9taXVtXSBpZiBrZXlkb3duIGlzIHByZXZlbnRlZCwgZG9uJ3QgdXBkYXRlIHRoZSBJTUUgYW5k IGNsZWFyIHRoZSBJTUUgc3RhdGUKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hv d19idWcuY2dpP2lkPTM5NDc1CisKKyAgICAgICAgKiBzcmMvV2ViVmlld0ltcGwuY3BwOgorICAg ICAgICAoV2ViS2l0OjpXZWJWaWV3SW1wbDo6aGFuZGxlQ29tcG9zaXRpb25FdmVudCk6CisKIDIw MTAtMDUtMjAgIFZpZXQtVHJ1bmcgTHV1ICA8dmlldHRydW5nbHV1QGNocm9taXVtLm9yZz4KIAog ICAgICAgICBSZXZpZXdlZCBieSBEYXJpbiBGaXNoZXIuCmRpZmYgLS1naXQgYS9XZWJLaXQvY2hy b21pdW0vc3JjL1dlYlZpZXdJbXBsLmNwcCBiL1dlYktpdC9jaHJvbWl1bS9zcmMvV2ViVmlld0lt cGwuY3BwCmluZGV4IDA2NGY4MDFjYmE4MzAwZDVkMmI4YjlmOTRiMmRlYTQ0Nzk3Yzk3ZjguLmMx NzBkOGE1MGJmMDEwYjA5YTM5MGM4ZmI1MzYzMjU3NTZhNWQ2YmIgMTAwNjQ0Ci0tLSBhL1dlYktp dC9jaHJvbWl1bS9zcmMvV2ViVmlld0ltcGwuY3BwCisrKyBiL1dlYktpdC9jaHJvbWl1bS9zcmMv V2ViVmlld0ltcGwuY3BwCkBAIC0xMTY0LDkgKzExNjQsNyBAQCBib29sIFdlYlZpZXdJbXBsOjpo YW5kbGVDb21wb3NpdGlvbkV2ZW50KFdlYkNvbXBvc2l0aW9uQ29tbWFuZCBjb21tYW5kLAogICAg ICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICBjb25zdCBXZWJTdHJpbmcmIGlt ZVN0cmluZykKIHsKICAgICBGcmFtZSogZm9jdXNlZCA9IGZvY3VzZWRXZWJDb3JlRnJhbWUoKTsK LSAgICAvLyBJZiB3ZSdyZSBub3QgZ29pbmcgdG8gZmlyZSBhIGtleXByZXNzIGV2ZW50LCB0aGVu IHRoZSBrZXlkb3duIGV2ZW50IHdhcwotICAgIC8vIGNhbmNlbGVkIHNvIGRvbid0IHVwZGF0ZSB0 aGUgSU1FLgotICAgIGlmICghZm9jdXNlZCB8fCAhbV9pbWVBY2NlcHRFdmVudHMgfHwgbV9zdXBw cmVzc05leHRLZXlwcmVzc0V2ZW50KQorICAgIGlmICghZm9jdXNlZCB8fCAhbV9pbWVBY2NlcHRF dmVudHMpCiAgICAgICAgIHJldHVybiBmYWxzZTsKICAgICBFZGl0b3IqIGVkaXRvciA9IGZvY3Vz ZWQtPmVkaXRvcigpOwogICAgIGlmICghZWRpdG9yKQpAQCAtMTE5MCw3ICsxMTg4LDkgQEAgYm9v bCBXZWJWaWV3SW1wbDo6aGFuZGxlQ29tcG9zaXRpb25FdmVudChXZWJDb21wb3NpdGlvbkNvbW1h bmQgY29tbWFuZCwKICAgICAgICAgICAgIHJldHVybiBmYWxzZTsKICAgICB9CiAKLSAgICBpZiAo Y29tbWFuZCA9PSBXZWJDb21wb3NpdGlvbkNvbW1hbmREaXNjYXJkKSB7CisgICAgLy8gSWYgd2Un cmUgbm90IGdvaW5nIHRvIGZpcmUgYSBrZXlwcmVzcyBldmVudCwgdGhlbiB0aGUga2V5ZG93biBl dmVudCB3YXMKKyAgICAvLyBjYW5jZWxlZC4gIEluIHRoYXQgY2FzZSwgY2FuY2VsIGFueSBleGlz dGluZyBjb21wb3NpdGlvbi4KKyAgICBpZiAoKGNvbW1hbmQgPT0gV2ViQ29tcG9zaXRpb25Db21t YW5kRGlzY2FyZCkgfHwgbV9zdXBwcmVzc05leHRLZXlwcmVzc0V2ZW50KSB7CiAgICAgICAgIC8v IEEgYnJvd3NlciBwcm9jZXNzIHNlbnQgYW4gSVBDIG1lc3NhZ2Ugd2hpY2ggZG9lcyBub3QgY29u dGFpbiBhIHZhbGlkCiAgICAgICAgIC8vIHN0cmluZywgd2hpY2ggbWVhbnMgYW4gb25nb2luZyBj b21wb3NpdGlvbiBoYXMgYmVlbiBjYW5jZWxlZC4KICAgICAgICAgLy8gSWYgdGhlIG9uZ29pbmcg Y29tcG9zaXRpb24gaGFzIGJlZW4gY2FuY2VsZWQsIHJlcGxhY2UgdGhlIG9uZ29pbmcK