From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020830 Description of problem: gedit has problems with three standard gb18030 testing files, `cat` them in gnome-terminal is ok. 1. double1.txt, there is one character whose GB code is A9F0(UxE801), it should appear as "unicode squre" in gedit, but right now a strange character looks like a Tibetan. The position is the 1st character of the last line. 2. user1.txt, this file should be totally unicode squares in gedit, but right now, the first 2 paragraphs have strange characters and spaces. 3. wei.txt, strange wrong bold characters are randomly appeared. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1.GinGin-re0128.nightly 2.Open those three testing files from here: http://junk.brisbane.redhat.com/gb18030/ 3. Additional info:
Sorry, about double1.txt, it should be 3 characters, not one, although they are looked like one large. the unicodes are from UxE801-UxE803.
Owen, is it because we have some new fonts that covers these codepoints? gnome-terminal will be unified across the screen because it uses one font across the way.
Is this gedit-specific or does it happen in any GtkTextView widget?
it happens in any GtkTextView widget
This is a blocker for our GB18030 compliance test so raising priority. To meet the current GM schedule we must ship CDs for testing to the PRC by Thursday.
I assume user1.txt is PUA characters. there is no definition of these at all, so why do you expect them to show as Unicode hex squares? Yes, the values presumably come from some font on the system which has some glyphs in the PUA... if you strace -e open, you should be able to figure out which one fairly easily.
Yeah, user1.txt is in PUA. When testing centre opens the user1.txt, because there is no corresponding glyph(in GB18030), so they wanted to see full unicode squares like user2 and user3.
Created attachment 89987 [details] Screenshot of the problem part of double1.txt
Created attachment 89988 [details] another problem part in double2.txt This is another problem that eng-asia discovered just then, for double1.txt it looks like some hebrew, but for double2.txt, it looks something like arabic characters.
double1.txt and double2.txt are just like user1.txt -- since these codes are not assigned in GB18030, they aren't in the font, fontconfig displays whatever font in the system does contain them, and some of our hebrew and arabic fonts have miscellaneous glyphs at various points in the PUA space. I don't see how this is incorrect behavior. But can't you remove fonts-hebrew and fonts-arabic for GB18030 compliance testing? The characters appearing in Wei.txt are for various reasons: - In some cases Zysong doesn't contain assigned Unicode code points (such as Arabic digits or Arabic presentation forms) - In other cases, some of the fonts on the system contain glyphs at unassigned code points for one reason or the other In no cases does it appear that characters that are present in Zysong.ttf are appearingly in a different font. [I'd really like to see us in the future shipping a GB18030 font without the "copied from the Unicode book" glyphs for Arabic and other similar languages. It causes considerable problems to have a font that pretends to cover languages, but doesn't really cover them.]
Jeremy, can we exclude the those font packages in comps if it is zh_CN?
The font pkgs that we have problem with are: double1.txt - fonts-hebrew double2.txt - fonts-arabic user1.txt - bitmap-fonts I have uninstalled those pkgs and tested the files and display is 'fine' after that.
Closing it for now.