Bug 86369 – [EFL] Try to use FontPango.cpp for international characters.

• Component: WebKit EFL
• Status: RESOLVED WONTFIX

[Ryuan Choi]

WebKit/Efl use freetype for default font backend like Gtk+.

But, FontPango.cpp can be used for international characters.

[Ryuan Choi]

Created attachment 141725 [details]
first_suggestion

[Raphael Kubo da Costa (:rakuco)]

Comment on attachment 141725 [details]
first_suggestion

View in context: https://bugs.webkit.org/attachment.cgi?id=141725&action=review

Mixing different backends that way looks weird, are you sure that's going to work. I also see some checks for freetype in FontPango.cpp itself -- does it make sense to switch to always using pango and simply enable or disable freetype support?

> Source/WebCore/ChangeLog:8
> +        Although freetype backend is choosed, FontPango.cpp can be used for

Nit: s/choosed/chosen. The description also sounds like the freetype backend is always chosen. I'd rephrase it to something like "Use FontPango.cpp even if the freetype backend is chosen, so that international characters can be rendered correctly".

> Source/cmake/OptionsEfl.cmake:107
> -  FIND_PACKAGE(Pango REQUIRED)
> -  SET(WTF_USE_PANGO 1)
> -  ADD_DEFINITIONS(-DWTF_USE_PANGO=1)
> +    FIND_PACKAGE(Pango REQUIRED)
> +    SET(WTF_USE_PANGO 1)
> +    ADD_DEFINITIONS(-DWTF_USE_PANGO=1)

Even though the current indentation is wrong, I don't think you should change it in this patch -- it just makes navigating through history slightly more difficult.

[Raphael Kubo da Costa (:rakuco)]

CC'ing some of the GTK folks who might know better about whether it makes sense to mix freetype and pango.

[Martin Robinson]

I don't think you want to default to the Pango backend even if it is present on the system. The Pango backend is missing a lot of features and once we get HarfBuzz support I'd like to just delete it.

[Ryuan Choi]

(In reply to comment #2)
> (From update of attachment 141725 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=141725&action=review
> 
> Mixing different backends that way looks weird, are you sure that's going to work. I also see some checks for freetype in FontPango.cpp itself -- does it make sense to switch to always using pango and simply enable or disable freetype support?

Now, pango backend is poor to use and freetype can not render complex font.
So IIRC, WebKit/Gtk+ use freetype and FontPango.cpp.

But I agree that it looks something weird.
I think that we can always use freetype backend with FontPango.cpp, instead.

> 
> > Source/WebCore/ChangeLog:8
> > +        Although freetype backend is choosed, FontPango.cpp can be used for
> 
> Nit: s/choosed/chosen. The description also sounds like the freetype backend is always chosen. I'd rephrase it to something like "Use FontPango.cpp even if the freetype backend is chosen, so that international characters can be rendered correctly".

Thank you. I'll do that.

> 
> > Source/cmake/OptionsEfl.cmake:107
> > -  FIND_PACKAGE(Pango REQUIRED)
> > -  SET(WTF_USE_PANGO 1)
> > -  ADD_DEFINITIONS(-DWTF_USE_PANGO=1)
> > +    FIND_PACKAGE(Pango REQUIRED)
> > +    SET(WTF_USE_PANGO 1)
> > +    ADD_DEFINITIONS(-DWTF_USE_PANGO=1)
> 
> Even though the current indentation is wrong, I don't think you should change it in this patch -- it just makes navigating through history slightly more difficult.

OK, I'll let indentation like before.

(In reply to comment #4)
> I don't think you want to default to the Pango backend even if it is present on the system. The Pango backend is missing a lot of features and once we get HarfBuzz support I'd like to just delete it.

Yes, I was waiting harfbuzz.
But now, FontPango.cpp is usefull to solve complex font issues on WebKit/Efl until we can use harfbuzz.

I'll prepare second approach that removes pango backend of WebKit/Efl and uses FontPango.cpp if pango is available.

Thank you.

[Dominik Röttsches (drott)]

(In reply to comment #4)
> I don't think you want to default to the Pango backend even if it is present on the system. The Pango backend is missing a lot of features and once we get HarfBuzz support I'd like to just delete it.

Is HarfBuzz support tracked in another bug? What's the status there - could we help with completing the implementation?

[Ryuan Choi]

Created attachment 145008 [details]
first suggestion(updated)

[Ryuan Choi]

(In reply to comment #7)
> Created an attachment (id=145008) [details]
> first suggestion(updated)

Updated rakuco's comment.

At least, tizen use this approach and it makes fast/text/international and some test cases unskipped.

[Martin Robinson]

So, if Tizen is using Pango as the backend, I guess that means that web fonts are broken or is that fixed there as well?

[Ryuan Choi]

(In reply to comment #9)
> So, if Tizen is using Pango as the backend, I guess that means that web fonts are broken or is that fixed there as well?

I did not receive any reports about it for a long time.

I will check and let you know.

[Ryuan Choi]

After r123864, efl port uses harfbuzz so that it is not needed anymore.