Some Japanese text style missing in Webpositive

[mt]

Since hrev54299, some Japanese text style is missing in Webpositive.

Related: ticket#16189

To reproduce:

1. Set Japanese fonts in Apparence.
2. Set Japanese fonts in Webpositive settings.
3. Open Japanese version of "Welcome to Haiku!" (or Quick tour, Userguide) in Webpositive.

[mt]

Webpositive opening Japanese version of "Welcome to Haiku!" in hrev54299.

[nephele]

I'm working on this, underlying issue is that noto sans bold fallback does not fall back to the correct style and instead falls back to style 0, which appears to be noto sans cjk thin in this case

[bitigchi]

​https://review.haiku-os.org/c/haiku/+/2887

[mt]

Replying to bitigchi:

​https://review.haiku-os.org/c/haiku/+/2887

Thanks for patch, it looks better.

[mt]

Webpositive opening Japanese version of "Welcome to Haiku!" in hrev54302 + patch.

[nephele]

The patch makes it fall back to Noto Sans CJK Regular instead of Noto Sans CJK Thin now, it's not the complete solution though, ideally it should fall back to Noto Sans CJK Bold instead :)

[mt]

I try to add bold check...

[nephele]

fontStyle = font.Style(); is the iteration that should set bold already if bold is set before, you can see that the fallback works for Noto Sans Display Bold in stylededit for example, does your patch make it work for Noto Sans Bold? (there would be other styles to check aswell)

[nephele]

Looks like this patch makes bold -> bold and bold italic -> bold work

[waddlesplash]

The "face" flag can be passed through to the fallback glyph selector, I believe, and perhaps this should be done as part of iterating.

[nephele]

It looks like face and Style serve a similar purpose, what is the difference between them? I suppose we should use one or the other

[mt]

Replying to nephele:

fontStyle = font.Style(); is the iteration that should set bold already if bold is set before, you can see that the fallback works for Noto Sans Display Bold in stylededit for example, does your patch make it work for Noto Sans Bold? (there would be other styles to check aswell)

I think we need check what font.Style() returns. If font.Style() returns simply "Bold", we could fallback "Noto Sans CJK JP Bold" already.

[mt]

Replying to nephele:

Looks like this patch makes bold -> bold and bold italic -> bold work

edit: that patch as is would also set bold as style unconditionally if bold italic is present for fonts that do have bold italic so it's not a perfect solution

Fallback fonts Noto Sans Thai, CJK JP, and Symbol don't have "Italic". So I think we check "Bold" first, but for "Noto Sans display" fallback, we check "Bold Italic" first...

[nephele]

I think we need check what font.Style() returns. If font.Style() returns simply "Bold", we could fallback "Noto Sans CJK JP Bold" already.

This already happens, I am just not sure why it does not work if it falls back from Noto sans bold as opposed to noto sans display bold (It might be a simple problem, but i have been really unsucsesfull to get debug output from thsi part of the code :/)

Fallback fonts Noto Sans Thai, CJK JP, and Symbol don't have "Italic". So I think we check "Bold" first, but for "Noto Sans display" fallback, we check "Bold Italic" first...

Problem is that this code should later use any user installed fonts to fallback to aswell, and we can't make many assumptions about which styles are available or not

I think the best way is to

1. match style
2. match style, strip italic part of style
3. try "Regular" style
4. try any style

the code currently does 1. and 4., with 1. not working correctly always (and 3. beeing the patch on gerrit)

[mt]

Patch 2

[mt]

I add Italic and Bold Italic support and stripping Italic from Noto Sans Thai, CJK JP, and Symbol.

[nephele]

This does work on haiku master somewhat, fallback is broken for thin italic though, and likely other *thin variants aswell, surely there is a better aproach to strip italic that does not require a hardcoded has italics list?

I should investiage more whether i can't get noto sans bold to fallback normally in the first iteration :)

[mt]

Fixed in hrev54747, thanks!!