Description of problem: Running firefox with pango enabled, I see ASCII text in Japanese rendered badly at the wrong size and with a bitmap font that makes it stand out. This is on a FC3 box. Version-Release number of selected component (if applicable): firefox-1.0-1 How reproducible: every time Steps to Reproduce: 1. Start a new firefox session with: $ LANG=ja_JP.UTF-8 firefox "http://headlines.yahoo.co.jp/hl?a=20041111-00000202-yom-bus_all" 2. Start a new firefox session with: $ MOZ_ENABLE_PANGO=1 LANG=ja_JP.UTF-8 firefox "http://headlines.yahoo.co.jp/hl?a=20041111-00000202-yom-bus_all" Actual results: (1) renders well with ASCII in the same font as the Japanese text as expected. (2) renders with ASCII in an undersized bitmap font which looks bad. Expected results: (2) Should look pretty much like (1) with the same anti-aliased font for ASCII text.
Created attachment 106472 [details] firefox-ja-no-pango-yahoo.png Shot of (1) without pango. Notice how "FRB" etc is in the right size and font.
Created attachment 106473 [details] firefox-ja-pango-yahoo.png With pango (2). Notice how "FRB" in the title is tiny and the other ascii text in the text too. Actually the highlighted bracketed text moves around when it is highlighted - though it can't be seen in this this screenshot. The openning "(" is actually rendered correctly, but the "FRB)" is small: when highlighted this text moves around.
Try install Japanese translation for firefox. There is a similar bug for Mozilla: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=107617
Pango seems to fix the problem in bug 107617. :) The problem reported here is a problem caused by pango rendering.
Reassigning to pango. You can see this behavior in other pango using applications. One example: 1. xchat 2. enable cannaLE 3. Type "The" 4. Hit SPACE ENTER 5. Type "TESThe" 6. hit SPACE ENTER Result: Ｔ㸠has the "T" is smaller than 㸠because it is an ugly bitmap font. TEST㸠has "TEST" the same size as ã¸.
Oops, the test case in Comment #5 above is bad. However this is still a pango or font problem where multibyte latin characters are falling back to the bitmap fonts.
Those two test cases are just the way CannaLE outputs. The real question is: the multibyte 'T' (ï¼´) is using Fixed seems, and Kochi has that glyphs (U+FF34)
Owen, seems pango is using Fixed for the multibyte 'T' (ï¼´). Is it a known behavior on pango? (as mentioned, Kochi has that glyphs (U+FF34))
This is basicallya dup of bug 107952. Pango says 'T', that's Latin, not Han, this 'ja' language tag doens't make sense, let me tag this with 'en'. But fontconfig doesn't consider that Kochi is a good font for the 'en' language tag. Probably easiest fix is to strip down the 'en' coverage that fontconfig requires to not have all extended characters. A fix for this specifically would be to reassign wide latin characters to be Han instead of Latin, or to make them script neutral. But that would cause Pango's script assigmments to deviate from the Unicode standard, which I don't like. *** This bug has been marked as a duplicate of 107952 ***
I just tried fontconfig-2.2.99 and it seems to fix this issue too. :) Will it be in fc4?
Changed to 'CLOSED' state since 'RESOLVED' has been deprecated.