49868 2010-11-20 12:47:57 -0800 REGRESSION (r71934): Main search field at Apple tech specs does not accept typing 2010-11-29 14:14:53 -0800 1 1 1 Unclassified WebKit DOM 528+ (Nightly build) Mac OS X 10.6 RESOLVED FIXED http://support.apple.com/specs/ InRadar, NeedsReduction, Regression P1 Major --- 46015 1 mitz dglazkov ap darin dglazkov hyatt pfeldman oldest_to_newest 311969 0 mitz 2010-11-20 12:47:57 -0800 To reproduce, navigate to the URL, focus the search field next to “Search Tech Specs” and type. 311970 1 mitz 2010-11-20 12:49:12 -0800 <rdar://problem/8692047> 311987 2 ap 2010-11-20 14:15:00 -0800 The only relevant commit in this range is the shadow DOM-aware event targeting one. 312008 3 mrowe 2010-11-20 15:42:12 -0800 Confirmed via bisection that <http://trac.webkit.org/changeset/71934> is to blame. This also breaks the search field in the web inspector. 312010 4 hyatt 2010-11-20 16:15:54 -0800 Just a guess, but focus/blur events that remain inside the shadow nodes of the same enclosing explicit element should not retarget. The same is also true for mouseover and mouseout events that remain inside the object. If those are now propagating out, that would cause odd behaviors. See: http://www.w3.org/TR/xbl/#event2 and then also sections 6.10 and 6.11. 312108 5 pfeldman 2010-11-21 07:35:44 -0800 Is anyone looking at this? It does break inspector's search. 312361 6 dglazkov 2010-11-22 08:58:03 -0800 Fixing.. 312829 7 dglazkov 2010-11-23 07:20:11 -0800 (In reply to comment #5) > Is anyone looking at this? It does break inspector's search. FWIW, tech specs behavior is slightly this is different from inspector's search. You can still search in Inspector by clicking on the loupe. I've been tracking through the code, have some leads. 312830 8 dglazkov 2010-11-23 07:21:55 -0800 (In reply to comment #4) > Just a guess, but focus/blur events that remain inside the shadow nodes of the same enclosing explicit element should not retarget. The same is also true for mouseover and mouseout events that remain inside the object. If those are now propagating out, that would cause odd behaviors. > > See: > > http://www.w3.org/TR/xbl/#event2 > > and then also > > sections 6.10 and 6.11. Yup, I didn't yet implement that. However, all of the shadow DOM elements today are explicitly marked as not focusable. I think the issue is with the style recalc not being triggered when focusing somehow. If you go in Inspector and modify width, the problem goes away. 313529 9 74812 dglazkov 2010-11-24 16:50:44 -0800 Created attachment 74812 Patch 313533 10 dglazkov 2010-11-24 16:52:18 -0800 (In reply to comment #9) > Created an attachment (id=74812) [details] > Patch This corrects the issue with the site reported. The problem still persists with the Inspector, but it's a totally different issue, related to manipulating (removing, actually) selection while inside of the focus event. I'll file a separate bug. 313845 11 pfeldman 2010-11-25 12:55:31 -0800 > This corrects the issue with the site reported. The problem still persists with the Inspector, but it's a totally different issue, related to manipulating (removing, actually) selection while inside of the focus event. I'll file a separate bug. Any idea what regressed it? Is there a bug already? 313846 12 dglazkov 2010-11-25 13:12:47 -0800 (In reply to comment #11) > > This corrects the issue with the site reported. The problem still persists with the Inspector, but it's a totally different issue, related to manipulating (removing, actually) selection while inside of the focus event. I'll file a separate bug. > > Any idea what regressed it? Is there a bug already? No yet, I'll file and provide reduction and fix after the Thanksgiving break. It regressed with the same revision. 314159 13 74812 darin 2010-11-27 18:26:51 -0800 Comment on attachment 74812 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=74812&action=review I wonder if some of the code in RenderBlock::hasLineIfEmpty can now be deleted. > LayoutTests/fast/forms/disabled-search-input.html:59 > \ No newline at end of file We normally put those end-of-file newlines in WebKit source files. > WebCore/rendering/TextControlInnerElements.cpp:54 > + virtual bool hasLineIfEmpty() const { return true; } I slightly prefer that such overrides be private rather than public. It’s typically a programming error if someone calls the function on a RenderTextControlInnerBlock* directly, and it’s nice to have this caught at compile time. And the privacy has no effect on overriding. 314161 14 dglazkov 2010-11-27 18:42:38 -0800 (In reply to comment #13) > (From update of attachment 74812 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=74812&action=review > > I wonder if some of the code in RenderBlock::hasLineIfEmpty can now be deleted. Not yet. The spin buttons still use it :( But soon, soon, I shall make this all be not so crappy! :) > > > LayoutTests/fast/forms/disabled-search-input.html:59 > > \ No newline at end of file > > We normally put those end-of-file newlines in WebKit source files. > > > WebCore/rendering/TextControlInnerElements.cpp:54 > > + virtual bool hasLineIfEmpty() const { return true; } > > I slightly prefer that such overrides be private rather than public. It’s typically a programming error if someone calls the function on a RenderTextControlInnerBlock* directly, and it’s nice to have this caught at compile time. And the privacy has no effect on overriding. Sounds great! Wil do. 314164 15 dglazkov 2010-11-27 19:49:08 -0800 (In reply to comment #14) > (In reply to comment #13) > > (From update of attachment 74812 [details] [details]) > > View in context: https://bugs.webkit.org/attachment.cgi?id=74812&action=review > > > > I wonder if some of the code in RenderBlock::hasLineIfEmpty can now be deleted. > > Not yet. The spin buttons still use it :( But soon, soon, I shall make this all be not so crappy! :) I think I can remove one check though. I'll do that. > > > > > > LayoutTests/fast/forms/disabled-search-input.html:59 > > > \ No newline at end of file > > > > We normally put those end-of-file newlines in WebKit source files. > > > > > WebCore/rendering/TextControlInnerElements.cpp:54 > > > + virtual bool hasLineIfEmpty() const { return true; } > > > > I slightly prefer that such overrides be private rather than public. It’s typically a programming error if someone calls the function on a RenderTextControlInnerBlock* directly, and it’s nice to have this caught at compile time. And the privacy has no effect on overriding. > > Sounds great! Wil do. 314474 16 dglazkov 2010-11-29 09:52:31 -0800 Committed r72807: <http://trac.webkit.org/changeset/72807> 74812 2010-11-24 16:50:44 -0800 2010-11-27 18:26:51 -0800 Patch bug-49868-20101124165043.patch text/plain 4293 dglazkov ZGlmZiAtLWdpdCBhL0xheW91dFRlc3RzL0NoYW5nZUxvZyBiL0xheW91dFRlc3RzL0NoYW5nZUxv ZwppbmRleCBjYWYyNDc3YjQzNWE5ZWNmZjcyYzJjOTY3M2Q5MDJiZjEyMGQzYTc1Li4zZDI4OWJj ZTZkNjZkMGI5ZDUyZGYwY2E0MDZjOTUzYmQwNDM5NTU4IDEwMDY0NAotLS0gYS9MYXlvdXRUZXN0 cy9DaGFuZ2VMb2cKKysrIGIvTGF5b3V0VGVzdHMvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTMgQEAK KzIwMTAtMTEtMjQgIERpbWl0cmkgR2xhemtvdiAgPGRnbGF6a292QGNocm9taXVtLm9yZz4KKwor ICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBSRUdSRVNTSU9O IChyNzE5MzQpOiBNYWluIHNlYXJjaCBmaWVsZCBhdCBBcHBsZSB0ZWNoIHNwZWNzIGRvZXMgbm90 IGFjY2VwdCB0eXBpbmcKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcu Y2dpP2lkPTQ5ODY4CisKKyAgICAgICAgKiBmYXN0L2Zvcm1zL2Rpc2FibGVkLXNlYXJjaC1pbnB1 dC1leHBlY3RlZC50eHQ6IEFkZGVkLgorICAgICAgICAqIGZhc3QvZm9ybXMvZGlzYWJsZWQtc2Vh cmNoLWlucHV0Lmh0bWw6IEFkZGVkLgorCiAyMDEwLTExLTIzICBJbHlhIFRpa2hvbm92c2t5ICA8 bG9pc2xvQGNocm9taXVtLm9yZz4KIAogICAgICAgICBVbnJldmlld2VkLiBVcGRhdGVkIGNocm9t aXVtIGV4cGVjdGF0aW9ucy4KZGlmZiAtLWdpdCBhL0xheW91dFRlc3RzL2Zhc3QvZm9ybXMvZGlz YWJsZWQtc2VhcmNoLWlucHV0LWV4cGVjdGVkLnR4dCBiL0xheW91dFRlc3RzL2Zhc3QvZm9ybXMv ZGlzYWJsZWQtc2VhcmNoLWlucHV0LWV4cGVjdGVkLnR4dApuZXcgZmlsZSBtb2RlIDEwMDY0NApp bmRleCAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwLi5mNTE3NWFiNDA5 YzEzM2EwOGVjMTgwZmQyZTc4YjRmMTgzMzBkOWNjCi0tLSAvZGV2L251bGwKKysrIGIvTGF5b3V0 VGVzdHMvZmFzdC9mb3Jtcy9kaXNhYmxlZC1zZWFyY2gtaW5wdXQtZXhwZWN0ZWQudHh0CkBAIC0w LDAgKzEsMiBAQAorCitUaGUgaW5uZXIgdGV4dCBlbGVtZW50IG9mIHRoZSBzZWFyY2ggaW5wdXQg c2hvdWxkIG5ldmVyIGJlIDAtaGVpZ2h0OiBQQVNTCmRpZmYgLS1naXQgYS9MYXlvdXRUZXN0cy9m YXN0L2Zvcm1zL2Rpc2FibGVkLXNlYXJjaC1pbnB1dC5odG1sIGIvTGF5b3V0VGVzdHMvZmFzdC9m b3Jtcy9kaXNhYmxlZC1zZWFyY2gtaW5wdXQuaHRtbApuZXcgZmlsZSBtb2RlIDEwMDY0NAppbmRl eCAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwLi41NmYzZmVlZTFmMmMz NTJkMDk4Yjk1N2JhYjc3ZDc1ODljMGMwOWQwCi0tLSAvZGV2L251bGwKKysrIGIvTGF5b3V0VGVz dHMvZmFzdC9mb3Jtcy9kaXNhYmxlZC1zZWFyY2gtaW5wdXQuaHRtbApAQCAtMCwwICsxLDU4IEBA Cis8aHRtbD4KKzxoZWFkPgorPHNjcmlwdD4KKworaWYgKHdpbmRvdy5sYXlvdXRUZXN0Q29udHJv bGxlcikgeworICAgIGxheW91dFRlc3RDb250cm9sbGVyLmR1bXBBc1RleHQoKTsKKyAgICBsYXlv dXRUZXN0Q29udHJvbGxlci53YWl0VW50aWxEb25lKCk7Cit9CisKK3ZhciBsb2dEaXY7CisKK2Z1 bmN0aW9uIGxvZyhtc2csIHN1Y2Nlc3MpCit7CisgICAgbG9nRGl2LmFwcGVuZENoaWxkKGRvY3Vt ZW50LmNyZWF0ZUVsZW1lbnQoJ2RpdicpKS50ZXh0Q29udGVudCA9IG1zZyArICc6ICcgKyAoISFz dWNjZXNzID8gJ1BBU1MnIDogJ0ZBSUwnKTsKK30KKworZnVuY3Rpb24gY2xpY2tPbihlbGVtZW50 KQoreworICAgIGlmICghd2luZG93LmV2ZW50U2VuZGVyKQorICAgICAgICByZXR1cm47CisKKyAg ICB2YXIgeCA9IGVsZW1lbnQub2Zmc2V0TGVmdCArIGVsZW1lbnQub2Zmc2V0V2lkdGggLyAyOwor ICAgIHZhciB5ID0gZWxlbWVudC5vZmZzZXRUb3AgKyBlbGVtZW50Lm9mZnNldEhlaWdodCAvIDI7 CisgICAgZXZlbnRTZW5kZXIubW91c2VNb3ZlVG8oeCwgeSk7CisgICAgZXZlbnRTZW5kZXIubW91 c2VEb3duKCk7CisgICAgZXZlbnRTZW5kZXIubW91c2VVcCgpOworfQorCitmdW5jdGlvbiBydW5U ZXN0KCkKK3sKKyAgICBsb2dEaXYgPSBkb2N1bWVudC5nZXRFbGVtZW50QnlJZCgnY29uc29sZScp OworICAgIHZhciBpbnB1dCA9IGRvY3VtZW50LmdldEVsZW1lbnRzQnlUYWdOYW1lKCdpbnB1dCcp WzBdOworICAgIHNldFRpbWVvdXQoZnVuY3Rpb24oKSB7CisgICAgICAgIGlucHV0LmRpc2FibGVk ID0gZmFsc2U7CisgICAgICAgIGlmICghd2luZG93LmV2ZW50U2VuZGVyKQorICAgICAgICAgICAg cmV0dXJuOworCisgICAgICAgIHZhciB4ID0gaW5wdXQub2Zmc2V0TGVmdCArIGlucHV0Lm9mZnNl dFdpZHRoIC8gMjsKKyAgICAgICAgdmFyIHkgPSBpbnB1dC5vZmZzZXRUb3AgKyBpbnB1dC5vZmZz ZXRIZWlnaHQgLyAyOworICAgICAgICBldmVudFNlbmRlci5tb3VzZU1vdmVUbyh4LCB5KTsKKyAg ICAgICAgZXZlbnRTZW5kZXIubW91c2VEb3duKCk7CisgICAgICAgIGV2ZW50U2VuZGVyLm1vdXNl VXAoKTsKKyAgICAgICAgZXZlbnRTZW5kZXIua2V5RG93bignYScpOworICAgICAgICBsb2coJ1Ro ZSBpbm5lciB0ZXh0IGVsZW1lbnQgb2YgdGhlIHNlYXJjaCBpbnB1dCBzaG91bGQgbmV2ZXIgYmUg MC1oZWlnaHQnLCBpbnB1dC52YWx1ZSA9PSAnYScpOworICAgICAgICBsYXlvdXRUZXN0Q29udHJv bGxlci5ub3RpZnlEb25lKCk7CisgICAgfSwgMCk7Cit9CisKKzwvc2NyaXB0PgorPC9oZWFkPgor PGJvZHkgb25sb2FkPSJydW5UZXN0KCkiPgorICAgIDxkaXY+CisgICAgICAgIDxpbnB1dCBpZD0i Zm9vIiB0eXBlPSJzZWFyY2giIHJlc3VsdHM9IjAiIGRpc2FibGVkPgorICAgIDwvZGl2PgorPGRp diBpZD0iY29uc29sZSI+Cis8L2Rpdj4KKzwvYm9keT4KKzwvaHRtbD4KXCBObyBuZXdsaW5lIGF0 IGVuZCBvZiBmaWxlCmRpZmYgLS1naXQgYS9XZWJDb3JlL0NoYW5nZUxvZyBiL1dlYkNvcmUvQ2hh bmdlTG9nCmluZGV4IDA3M2I4ZDRjZjI2NDg2NDU5NTI5YTNjN2IzMGU1MTU2MmNkOGE3NDkuLjZk ZjU5Yjk4MDMxYmFiZDcyMTgwNzZiMDQzMjgyY2I5MjRiNjU4YWIgMTAwNjQ0Ci0tLSBhL1dlYkNv cmUvQ2hhbmdlTG9nCisrKyBiL1dlYkNvcmUvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTUgQEAKKzIw MTAtMTEtMjQgIERpbWl0cmkgR2xhemtvdiAgPGRnbGF6a292QGNocm9taXVtLm9yZz4KKworICAg ICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBSRUdSRVNTSU9OIChy NzE5MzQpOiBNYWluIHNlYXJjaCBmaWVsZCBhdCBBcHBsZSB0ZWNoIHNwZWNzIGRvZXMgbm90IGFj Y2VwdCB0eXBpbmcKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dp P2lkPTQ5ODY4CisKKyAgICAgICAgVGVzdDogZmFzdC9mb3Jtcy9kaXNhYmxlZC1zZWFyY2gtaW5w dXQuaHRtbAorCisgICAgICAgICogcmVuZGVyaW5nL1RleHRDb250cm9sSW5uZXJFbGVtZW50cy5j cHA6CisgICAgICAgIChXZWJDb3JlOjpSZW5kZXJUZXh0Q29udHJvbElubmVyQmxvY2s6Omhhc0xp bmVJZkVtcHR5KToKKwogMjAxMC0xMS0yMyAgQWRhbSBCZXJna3Zpc3QgIDxhZGFtLmJlcmdrdmlz dEBlcmljc3Nvbi5jb20+CiAKICAgICAgICAgUmV2aWV3ZWQgYnkgTWFydGluIFJvYmluc29uLgpk aWZmIC0tZ2l0IGEvV2ViQ29yZS9yZW5kZXJpbmcvVGV4dENvbnRyb2xJbm5lckVsZW1lbnRzLmNw cCBiL1dlYkNvcmUvcmVuZGVyaW5nL1RleHRDb250cm9sSW5uZXJFbGVtZW50cy5jcHAKaW5kZXgg YmE0YzdiYjg1MDY1N2FjZjk1Y2MzZDIwNmFkMzdmNTU2OTJkNjdhNi4uMTAxNTU5ZmY5ZTNhMWMx NzBhYWY4YjYyZWYyOTEzZmQxZTVmNjYwMyAxMDA2NDQKLS0tIGEvV2ViQ29yZS9yZW5kZXJpbmcv VGV4dENvbnRyb2xJbm5lckVsZW1lbnRzLmNwcAorKysgYi9XZWJDb3JlL3JlbmRlcmluZy9UZXh0 Q29udHJvbElubmVyRWxlbWVudHMuY3BwCkBAIC01MSw2ICs1MSw4IEBAIGNsYXNzIFJlbmRlclRl eHRDb250cm9sSW5uZXJCbG9jayA6IHB1YmxpYyBSZW5kZXJCbG9jayB7CiBwdWJsaWM6CiAgICAg UmVuZGVyVGV4dENvbnRyb2xJbm5lckJsb2NrKE5vZGUqIG5vZGUsIGJvb2wgaXNNdWx0aUxpbmUp IDogUmVuZGVyQmxvY2sobm9kZSksIG1fbXVsdGlMaW5lKGlzTXVsdGlMaW5lKSB7IH0KIAorICAg IHZpcnR1YWwgYm9vbCBoYXNMaW5lSWZFbXB0eSgpIGNvbnN0IHsgcmV0dXJuIHRydWU7IH0KKwog cHJpdmF0ZToKICAgICB2aXJ0dWFsIFZpc2libGVQb3NpdGlvbiBwb3NpdGlvbkZvclBvaW50KGNv bnN0IEludFBvaW50Jik7CiAK