2008690 - Bad clipping/width of large operators when italic correction is applied

Status: RESOLVED FIXED

(Core :: MathML, defect)

Description

[Frédéric Wang Nélar]

Attachment: largeop-rtl-mirroring-clipping.html

See the attached testcase, with large clockwise integrals drawn with Latin Modern Math and XITS Math.

In LTR mode, the width is too small (see the blue background).

In RTL mode this is even worse: the left part of the integrals are never shown.

Both are due to the following code that subtracts italic correct from the operator width:

https://searchfox.org/firefox-main/rev/33bba5cfe4a89dda0ee07fa9fbac578353713fd3/layout/mathml/nsMathMLChar.cpp#1021

This is legacy stuff that mimics the behavior of old fonts like STIX-Word, that provides a shorter width for slanted integrals, in order to allow proper position of subscripts.

Per MathML Core, we should actually not modified the operator width but instead take the italic correction of the base during the layout of nsMathMLmmultiscriptsFrame: https://www.w3.org/TR/mathml-core/#base-with-subscript

Note that this is a bit tricky because the base may be more complex (e.g. <mrow><mrow><mrow><mo largeop="true">∲</mo></mrow></mrow></mrow>) and we need to transmit the italic correction from the core operator to the nsMathMLmmultiscriptsFrame, which is probably why we haven't done that so far.

I see:
https://www.w3.org/TR/mathml-core/#layout-of-mtext

If the text content is made of a single glyph and this glyph has an entry in the MathItalicsCorrectionInfo table then the specified value is used as the italic correction.

https://www.w3.org/TR/mathml-core/#layout-of-mrow

The italic correction of the math content is set to the italic correction of the last in-flow child, which is the final value of previous-italic-correction.

https://www.w3.org/TR/mathml-core/#layout-of-operators
If the algorithm to shape a stretchy glyph has been used for one of the step above, then the italic correction of the math content is set to the value returned by that algorithm.

https://wpt.fyi/results/mathml/presentation-markup/scripts/subsup-parameters-2.html seems to be a basic test for italic correction, but we probably need more complex ones involving mtext (or mo behaving like mtext) or mrow constructions.

Comment 1

[Frédéric Wang Nélar]

Attachment: Screenshot in Firefox release 146

Comment 2

[Frédéric Wang Nélar]

Attachment: Screenshot in nightly, with italic correction disabled.

Comment 3

[Frédéric Wang Nélar]

Attachment: Screenshot in Firefox release 146

Comment 4

[Frédéric Wang Nélar]

I attached one screenshot showing the bug in Firefox release 146. And another one showing the script placement issue if we disable the italic correction hack at https://searchfox.org/firefox-main/rev/33bba5cfe4a89dda0ee07fa9fbac578353713fd3/layout/mathml/nsMathMLChar.cpp#1021

Comment 5

[Frédéric Wang Nélar]

Attachment: Bug 2008690 - Make nsMathMLmmultiscriptsFrame handle italic correction of largeop base. r=emilio,eri

Comment 6

[Frédéric Wang Nélar]

https://treeherder.mozilla.org/jobs?repo=try&revision=297ecb3266a99ddd81a5f190f6dd0288142d530d

Comment 7

[Frédéric Wang Nélar]

https://treeherder.mozilla.org/jobs?repo=try&revision=89f9f6bb3f8c7fba5dc5e701451f39f726d5237e

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/12f92743d12b
https://hg.mozilla.org/integration/autoland/rev/cb43e719cd12
Make nsMathMLmmultiscriptsFrame handle italic correction of largeop base. r=emilio,layout-reviewers

Comment 9

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Created web-platform-tests PR https://github.com/web-platform-tests/wpt/pull/57335 for changes under testing/web-platform/tests

Comment 10

[Cristina Horotan [:chorotan]]

https://hg.mozilla.org/mozilla-central/rev/cb43e719cd12

Comment 11

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Upstream PR merged by moz-wptsync-bot

Comment 12

[Frédéric Wang Nélar]

Attachment: Latest nightly with the fix

Comment 13

[Hamish Willee]

FF149 MDN work for this can be tracked in https://github.com/mdn/content/issues/43207

Fred, what were you thinking was needed for this? It sounds like a bug fix, so perhaps:

1. A release note like this? (if not, please suggest alternative):

   Fixed a rendering bug where large operators were incorrectly clipped or had improper widths when italic correction was applied.

2. Does this need data compatibility information to be recorded? If so, is https://developer.mozilla.org/en-US/docs/Web/MathML/Reference/Element/mo#largeop the correct operator.attribute to target?

Comment 14

[Frédéric Wang Nélar]

@Hamish: Largeop has been supported forever in Firefox, I think this is more a bug fix so a note like you did in (1) sounds correct.

(I don't know how important it is to be in the release notes, but I'll trust your judgement on that)

Comment 15

[Hamish Willee]

Hi Fred, Thanks! I choose not to include anything for this - it doesn't change the way that developers should use the interface. I suspect most that care will aleady be tracking this issue, and the remainder hopefully upgrade to new Firefox and never notice there was ever a problem.