Add test checking the lang attribute affects rendering

[mbrodesser-Igalia]

The added test is self-describing.

[mbrodesser-Igalia]

The test passes for Chrome and Firefox. For WebKitGTK's MiniBrowser, it fails.

[drott]

Thanks for the test addition, this works, but I have some concerns about DejaVu Sans, in particular as this change suggests to add a second copy.

I noticed recently this font is not maintained since 2016 and also has some issues as missing cap height and x height metrics in the OS/2 table. Hence I would check to see if this could be demonstrated with open-source, more recently released and updated fonts, for example from the Noto collection or Roboto perhaps?

fonttools' ttx -tgsub -o- <font file> or HarfBuzz' hb-shape testing tool might help in identifying whether there are locale specific ligatures in such fonts.

[mbrodesser-Igalia]

Thanks for the test addition, this works, but I have some concerns about DejaVu Sans, in particular as this change suggests to add a second copy.

I noticed recently this font is not maintained since 2016 and also has some issues as missing cap height and x height metrics in the OS/2 table. Hence I would check to see if this could be demonstrated with open-source, more recently released and updated fonts, for example from the Noto collection or Roboto perhaps?

fonttools' ttx -tgsub -o- <font file> or HarfBuzz' hb-shape testing tool might help in identifying whether there are locale specific ligatures in such fonts.

Thanks for the pointers. Learning why the test currently passes with the DejaVu font, before replacing that font.

ttx is case-sensitive, i.e. ttx -tGSUB -o- <font-file> is required.

[jfkthame]

I'm happy to see additional test coverage for this sort of behavior. Just a note in passing: there's already a file css/css-fonts/language-specific-01.html that aims to exercise some language-dependent behavior; however, it lacks a linked reference file, so I guess it doesn't actually get used unless one loads and inspects it manually.

(It uses a copy of Lato, found in the support/fonts dir, which has different behavior for "fi" in Turkish vs default.)

So perhaps it'd be good to create a reference case for that file? (In addition to the new test being added here.)

[mbrodesser-Igalia]

I'm happy to see additional test coverage for this sort of behavior. Just a note in passing: there's already a file css/css-fonts/language-specific-01.html that aims to exercise some language-dependent behavior; however, it lacks a linked reference file, so I guess it doesn't actually get used unless one loads and inspects it manually.

(It uses a copy of Lato, found in the support/fonts dir, which has different behavior for "fi" in Turkish vs default.)

So perhaps it'd be good to create a reference case for that file? (In addition to the new test being added here.)

Thanks for the pointers.

Updated this PR.

I'll add a separate test, checking "fi" for English and Turkish are rendered differently for the Lato font. Since the test of this PR relies on that assumption.

[mbrodesser-Igalia]

I'll simplify the test of this PR. It's sufficient to check "fi" renders differently for Turkish and English.

[mbrodesser-Igalia]

I'm happy to see additional test coverage for this sort of behavior. Just a note in passing: there's already a file css/css-fonts/language-specific-01.html that aims to exercise some language-dependent behavior; however, it lacks a linked reference file, so I guess it doesn't actually get used unless one loads and inspects it manually.

(It uses a copy of Lato, found in the support/fonts dir, which has different behavior for "fi" in Turkish vs default.)

So perhaps it'd be good to create a reference case for that file? (In addition to the new test being added here.)

Will consider adding a reference case for that file, but in a separate PR.

Filed #50364 so that it's not forgotten, but I currently lack time for that change.

[drott]

LGTM, thanks for updating the test to a leaner version without requiring a font addition. I'd suggest to hear it from Jonathan on the updated PR as well.

[jfkthame]

LGTM