941607 - Incorrect preferred width of script l in <mi>

Status: RESOLVED FIXED

(Core :: MathML, defect, P4)

Description

[Frédéric Wang (:fredw)]

Attachment: testcase.html

Follow-up of bug 415413.

In the attach testcases, no linebreak should occur inside the table cells. With the patch of bug 415413, all but the first case are fixed. It seems that it has something to do with the fact that we are using an <mi> with a letter that is already italic. I'm curious to see the result after James' patch for bug 114365.

Comment 1

[Frédéric Wang (:fredw)]

Attachment: italic-l.png

Here is how it renders on Linux.

Comment 2

[Frédéric Wang (:fredw)]

(... after application of the patch from bug 415413)

Comment 3

[Frédéric Wang (:fredw)]

Attachment: Screenshot of Jacques' testcase with the mathvariant patches

Jacques' testcase (attachment 301094 [details]) looks much better with the mathvariant patches on Linux.

Comment 4

[Frédéric Wang (:fredw)]

Attachment: Screenshot of the testcase with the mathvariant patches

For the testcase, the last one still fails. I suspect there is something wrong in the italic correction or the width measurement ; where the italic l is used instead of the transformed script l.

Comment 5

[Frédéric Wang (:fredw)]

Attachment: testcase

There were invalid "font-family: italic" in the initial testcase. I fixed that and also force the use of STIX fonts so that we get consistent rendering.

Comment 6

[Frédéric Wang (:fredw)]

Attachment: Screenshot of the testcase with the mathvariant patches

Comment 7

[Frédéric Wang (:fredw)]

So attachment 301094 [details] and the testcase of this bug renders correctly for me on Windows and Linux with my latest try build.

@Jacques: can you try the Mac try build?
http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-51ab041e5ec1/try-macosx64/

Comment 8

[distler]

Attachment: Testcase in Frederic's try build

Better ...

Comment 9

[Frédéric Wang (:fredw)]

(In reply to distler from comment #8)
> Created attachment 8338999 [details]
> Testcase in Frederic's try build
> 
> Better ...

OK, I think the remaining failure is bug 941611, which seems platform dependent.

I'm guessing attachment 301094 [details] now works for you? BTW, I think the try build will also fix your bug 518592...

Comment 10

[Frédéric Wang (:fredw)]

Attachment: Patch (tests)

So this is fixed per the patch of bug 114365.

Comment 11

[Frédéric Wang (:fredw)]

Attachment: Patch (tests) - V2

Comment 12

[Karl Tomlinson (:karlt)]

Comment on attachment 8341917 [details] [diff] [review]
Patch (tests) - V2

>+fails-if((OSX&&(OSX!=10.8))||/^Windows\x20NT\x206\.[12]/.test(http.oscpu)||Android) == table-width-4.html table-width-4-ref.html

I'm not sure it makes sense to add this test yet, given the number of
platforms on which it fails, but perhaps someone will notice a failure on
Linux that can be reproduced, making this easier to fix.

Can (OSX<10.8) be used instead of (OSX!=10.8)?  There will be 10.9 tests soon, and 10.9 is more likely to be similar to 10.8 than older versions.
(I don't know whether OSX is an integer or string.)

Comment 13

[Karl Tomlinson (:karlt)]

Similarly, unless this already passes on windows 8.1, the corresponding check should try to include future versions.  Something like /^Windows\x20NT\x206\.[^0]/

Comment 14

[Frédéric Wang (:fredw)]

Attachment: Patch (tests) - V3

https://tbpl.mozilla.org/?tree=Try&rev=b13e85baf9a1

Comment 15

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/9b14d922a514

Comment 16

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/9b14d922a514