Overview
Request 755067 accepted
- Make sure at least one font with coverage for korean is installed,
and can be rendered by Cairo >= 1.44. Fixes boo#1158326.
- Created by StefanBruens
- In state accepted
Loading...
Request History
StefanBruens created request
- Make sure at least one font with coverage for korean is installed,
and can be rendered by Cairo >= 1.44. Fixes boo#1158326.
ftake accepted request
Thank you. I will add some comments after accept
I think it should not depend on specific fonts since we sometimes changes language's default fonts
It does not. "noto-sans-kr-fonts" is just a recommends. "scalable-font-ko" is a Provides available from several font packages.
Other fonts which fulfill the requirements are:
and any other from https://fontinfo.opensuse.org/blocks/block-HangulSyllables.html which is not a bitmap font (i.e. GNU Unifont).
This comment has been deleted
Currently, anyone doing a default install will end up with broken Hangul glyphs (i.e. just boxes), see the e.g. https://openqa.opensuse.org/tests/1102602#step/ibus_test_kr/18
Can we get this in, please? Or do something else to get this resolved ...
So, what happens if you remove "Recommends: noto-sans-kr-fonts" but keep only "Requires: scalable-font-ko"?
With recommends:
The following 5 NEW packages are going to be installed:
ibus-hangul libhangul1 noto-sans-kr-bold-fonts noto-sans-kr-fonts noto-sans-kr-regular-fonts
The following recommended package was automatically selected:
noto-sans-kr-fonts
Without recommends:
The following 3 NEW packages are going to be installed:
alee-fonts ibus-hangul libhangul1
I.e. zypper chooses the (alphabetically) first package which fits, i.e. it installs a semi-random font package.
And doesn't alee-fonts show any proper Koren glyph? If so, it implies that the Provides in alee-fonts is wrong, and we should get rid of it, no?
Depends on the definition of "proper" - the included fonts (5 different families) are apparently complete and likely correct.
But this does not mean alee-fonts is a good default choice - Noto Sans is IMHO much better designed, with proper hinting and the like.
The 5 fonts from alee-fonts: - https://fontinfo.opensuse.org/fonts/BandalRegular.html - https://fontinfo.opensuse.org/fonts/BangwoolRegular.html - https://fontinfo.opensuse.org/fonts/EunjinRegular.html - https://fontinfo.opensuse.org/fonts/EunjinNakseoRegular.html - https://fontinfo.opensuse.org/fonts/GuseulRegular.html
Noto Sans: https://fontinfo.opensuse.org/fonts/NotoSansKRRegular.html
What what are the reasons to not make Noto Sans the default? The user can choose any other font any time, it is only a weak dependency.
If you think there is a better default, I am open for suggestions. But we should choose one to get deterministic results. We should choose a font which renders nicely to avoid any 'asian fonts on SUSE look like shit' reports.
Unfortunately, what is difficult is that we don't have perfect Asian fonts so far. I am not sure which one is better.
So, what are you missing in Noto Sans? It has a 100% glyph coverage, it is well hinted, and readable even at small sizes. If you want something more fancy/artistic, have a look at Noto Serif.
The alee-fonts don't look very well balanced to me, so beyond good glyph coverage I don't think it is a good choice.
Second choice after Noto Sans and Serif would be Baekmuk Gulim and Baekmuk Batang, which look well balanced, though have less glyph coverage and are less readable at small sizes than Noto Sans.
Anyway, we can discuss the "perfect" font for some more weeks, while users end up with completely broken fonts. Or we can just use a sensible default, which can overridden locally any time.
The point is that it's your preference, and hard-coding your favorite in a random place is no good implementation in general.
OTOH, noto-fonts should be installed as default and enabled as the preferred font via fontconfig if user installs the system with ko locale. So, practically seen, using this font is no problem, per se.
Actually, the bug surfaced because zypper (or installer) doesn't pick up all matching fonts with scalable-ko-fonts but only the first alphabetic one (supposedly). If we fix this in the zypper / installer side, we don't need such a hack. But, it's a harder way and may bring other problems.
So, I think we can go for this way -- adding a Recommends noto-ko-fonts in this particular case. But, then a comment line is needed so that reader can understand why this specific one is put there, at least.
Should only one font package provide "scalable-ko-fonts"?
Noto Sans CJK KR is now default font for the ko locale (when LC_CTYPE is ko). https://github.com/openSUSE/fonts-config/blob/master/59-family-prefer-lang-specific.conf#L254
However, the font is not set as an alias sans-serif. So it might not be selected if some other Korean fonts are installed when LC_CTYPE is other than ko: https://github.com/openSUSE/fonts-config/blob/master/60-family-prefer.conf#L92
I agree going this way.
Definitely no - its a virtual provides which all suitable (scalable, glyph coverage) should have.
Does not help when you want to write Korean and your locale is different.
The system is not installed with ko locale but en, see the linked openQA run.
As ibus-hangul is installed later, no suitable font is available. People may want to write Korean even if their locale is different.
"Noto Sans has a good coverage" - isn't this sufficient?
IMO, the test scenario is insufficient. If you want to use Korean -- at least with the input method or such -- user should enable the Korean language support on the system, not only installing only ibus-hangul If you did it on YaST, all needed stuff should have been installed, including noto fonts.
But it's a bit off-topic, and having a local workaround is acceptable.
Shell we add more information in changelog? Or accept as is?
For further works, we need to make a new mechanism to install default font sets of specified language to resolve our concern.
I'd put more comment in both the spec and the changelog.
I will not invest any more time here - please go ahead and do whatever you deem necessary.
@MargueriteSu, @WernerFink, @fstrba, @ftake, @hillwood, @mrdocs, @swyear, @tiwai: review reminder