170229 2017-03-28 22:19:17 -0700 MathML: layout is unstable on page zoom 2018-01-13 04:17:43 -0800 1 1 1 Unclassified WebKit MathML WebKit Nightly Build Mac macOS 10.12 NEW P2 Normal --- 161126 180029 1 lambda webkit-unassigned alex fred.wang mmaxfield rego simon.fraser oldest_to_newest 1292303 0 305707 lambda 2017-03-28 22:19:17 -0700 Created attachment 305707 An example involving stretchy operator. In the given attachment, if one scales the webpage (via Command + or trackpad), then scales it back (via Command 0, or just press refresh), the layout of the equation changes. Scaling a page and scaling back should not change rendering results in any event. I am not familiar with WebKit’s structure; hopefully this bug does not imply some deep semantics issue around scaling in general and is specific to MathML only. The issue exists in Safari 10, TP, and nightly builds. Changing font from default old STIX one shipped with Mac to STIX two does not help. 1292742 1 fred.wang 2017-03-30 01:00:23 -0700 Thanks for the report. IIUC, this happens with stretchy operators (parenthesis here)? Maybe we can narrow down the report and reduce the test case? 1293180 2 305974 lambda 2017-03-31 09:35:11 -0700 Created attachment 305974 Minimal unstable example This examples involves only a stretchy operator (an triple integral symbol) and a identifier 'x'. It is also unstable on page zoom. 1293614 3 fred.wang 2017-04-03 02:08:46 -0700 Thanks for the reduced test case. I guess the code involved is the same as in bug 170272, so you should be able to debug and narrow down the issues? 1374861 4 fred.wang 2017-11-27 01:22:52 -0800 @Minsheng Were you able to work on this? Maybe you could start with bug 180029, which is somewhat simple to test. 1374982 5 lambda 2017-11-27 10:11:35 -0800 I am interested, but I might have only ten hours or less every week—still a full time student and the final week is approaching. I will take a serious look after we finish the stretchy case. 1389109 6 lambda 2018-01-13 04:17:43 -0800 It appears that the issue is gone, but I am not sure which patch resolves it. I suspect it is @fred’s previous refactoring of RenderMathMLRow which eliminates some dirty states. I plan to add a test as a part of my new proposal of combining the layout code for RenderMathMLRow, RenderMathMLEnclose, and RenderMathMLRoot. In any event, note that in the first example, the parentheses are a bit off. Something seems wrong with the vertical stretching. 305707 2017-03-28 22:19:17 -0700 2017-03-28 22:19:17 -0700 An example involving stretchy operator. int.html text/html 720 lambda PCFkb2N0eXBlIGh0bWw+CjxodG1sPgogIDxoZWFkPgogICAgPG1ldGEgY2hhcnNldD0idXRmLTgi PgogIDwvaGVhZD4KICA8Ym9keT4KICAgIDxwPlByZXNzIENvbW1hbmQgKyB0byBjaGFuZ2UgdGhl IHNjYWxlLCB0aGVuIHByZXNzIENvbW1hbmQgMCB0byByZXZlcnQKICAgIHRvIHRoZSBkZWZhdWx0 IHNjYWxlLiBQb3NpdGlvbiBvZiB4IHdvdWxkIGJlIGRpZmZlcmVudCBmcm9tIHdoYXQgaXQgd2Fz LgogICAgUmVmcmVzaCB0aGUgcGFnZSB0byBub3RlIHRoZSBkaWZmZXJlbmNlLjwvcD4KICAgIDxt YXRoIGRpc3BsYXk9ImJsb2NrIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMTk5OC9NYXRoL01h dGhNTCI+CiAgICAgIDxtcm93PgogICAgICAgIDxtc3Vic3VwPgogICAgICAgICAgPG1vPuKIrTwv bW8+CiAgICAgICAgICA8bW4+MDwvbW4+CiAgICAgICAgICA8bW4+MTwvbW4+CiAgICAgICAgPC9t c3Vic3VwPgogICAgICAgIDxtaT5mPC9taT4KICAgICAgICA8bXJvdz4KICAgICAgICAgIDxtbyBz dHJldGNoeT0idHJ1ZSIgZm9ybT0icHJlZml4Ij4oPC9tbz4KICAgICAgICAgIDxtaT54PC9taT4K ICAgICAgICAgIDxtbyBzdHJldGNoeT0idHJ1ZSIgZm9ybT0icG9zdGZpeCI+KTwvbW8+CiAgICAg ICAgPC9tcm93PgogICAgICAgIDxtaT5kPC9taT4KICAgICAgICA8bWk+eDwvbWk+CiAgICAgIDwv bXJvdz4KICAgIDwvbWF0aD4KICA8L2JvZHk+CjwvaHRtbD4K 305974 2017-03-31 09:35:11 -0700 2017-03-31 09:35:11 -0700 Minimal unstable example int2.html text/html 247 lambda PCFkb2N0eXBlIGh0bWw+CjxodG1sPgogIDxoZWFkPgogICAgPG1ldGEgY2hhcnNldD0idXRmLTgi PgogIDwvaGVhZD4KICA8Ym9keT4KICAgIDxtYXRoIGRpc3BsYXk9ImJsb2NrIiB4bWxucz0iaHR0 cDovL3d3dy53My5vcmcvMTk5OC9NYXRoL01hdGhNTCI+CiAgICAgIDxtcm93PgogICAgICAgIDxt byBzdHJldGNoeT0idHJ1ZSI+4oitPC9tbz4KICAgICAgICA8bWk+eDwvbWk+CiAgICA8L21hdGg+ CiAgPC9ib2R5Pgo8L2h0bWw+Cg==