STR 1. use latest m-c or m-i 2. type `m` in urlbar. my first result for `m` is https://mail.google.com/. 3. press `tab` to select first result url autocomplete. actual result - nothing happen, cannot use tab to select url autocomplete. expected result - can use tab to select url autocomplete. first bad revision is changeset: 92974:de5745bce8bc user: Masayuki Nakano <masayuki@d-toybox.com> date: Thu May 03 17:35:02 2012 +0900 files: content/events/test/test_dom_keyboard_event.html dom/base/nsDOMWindowUtils.cpp dom/interfaces/base/nsIDOMWindowUtils.idl layout/forms/test/test_bug348236.html testing/mochitest/tests/SimpleTest/EventUtils.js description: Bug 166240 part.8 Add tests for KeyboardEvent.location (synthesized events) r=smaug, sr=jst last good revision is changeset: 92966:a6a335cd2c94 user: Jonathan Kew <jkew@mozilla.com> date: Thu May 03 08:57:30 2012 +0100 files: layout/reftests/text-transform/greek-small-caps-1-ref.html layout/reftests/text-transform/greek-small-caps-1.html layout/reftests/text-transform/reftest.list description: bug 307039 - reftest for Greek small-caps behavior. r=past pushlog http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml/?fromchange=a6a335cd2c94&tochange=de5745bce8bc

Masayuki-san, could you please check/clarify how that fix affected the locationbar?

Hmm, I have no idea. In XP level, it shouldn't cause any regression. The widget part for Windows is a little bit complex. I guess that the cause should be in part.2.

I'd like someone to test this regression on other platforms.

Ah, I see. I found a logical bug in the previous code. But the behavior is broken by the new code.

Cannot reproduce on Linux. Mozilla/5.0 (X11; Linux i686; rv:15.0) Gecko/15.0 Firefox/15.0a1

Thanks. I think that the keypress events of ESC key and Tab key do not work.

testing on tryserver. https://tbpl.mozilla.org/?tree=Try&usebuildbot=1&rev=1d5540cfcde0

Comment on attachment 621077 [details] [diff] [review] Patch At OnChar(), it sets keycode from wParam of WM_CHAR when charCode is zero. It's logically wrong, but it worked because native virtual keycode and native charcode are same for Tab key and Esc key. This patch computes DOM keycode from scan code and sets it when keypress event's charCode is 0.