Closed Bug 89169 Opened 24 years ago Closed 24 years ago

meta charset doesn't work well on 07-03-11 trunk build

Categories

(Core :: Internationalization, defect)

x86
Windows NT
defect
Not set
normal

Tracking

()

VERIFIED FIXED

People

(Reporter: amyy, Assigned: vidur)

Details

(Keywords: intl, Whiteboard: verified on trunk)

Build: 07-03-11 Win32 trunk on WinNT. If there system doesn't have some non-ascii font(e.g. Japanese, Korean, Chinese...), then if you try to visit some pages that require those fonts, browser will pops-up a dialog for downloading. However, when I visit some CJK pages, there is no dialog pops-up, and I checked the charset were marked as Western(iso-8859-1) even they are has meta charset. Results: 1. http://home.netscape.com/ja, no matter auto-detect set to what, the charset will marked as shift-jis, which is the meta-charset in html source code. And with this page, it does pops-up a dialog for download the font. 2. http://home.netscape.com.ko, when auto-detect set to OFF or ALL, the charset will point to western(iso-8859-1), and there is no download font dialog pops-up. When auto-detect set to East Asian or Korean, then don't have this problem. However, with http://kr.yahoo.com, charset will point to EUC-KO and has the download dialog pops-up even auto-detect set to OFF or ALL. 3. http://home.netscape.com/zh/tw or http://home.netscape.com/zh/cn, got very similar results as netscape-korean page. And with some other chinese page like: http://www.chineseworld.com, charset will point to western(1252). You should create a new profile or clear the cache in order to reproduce it. If the system has those fonts, then you will not able to see this problem. Marina got similar result on her win2000 machine too.
Keywords: intl
QA Contact: andreasb → ylong
vidur, I think you should bake out your change to nsIParser.h yokoyama- it seems meta charset does not work neither.
reassign to vidur.
Assignee: nhotta → vidur
A patch for backing out the change that Frank mentions is below. I'm still not a fan of backing it out since it implies resniffing of meta tag information when we're coming out of the cache (with potential minor performance implications). On the other hand, if it fixes a set of auto-detection issues (which I admit I don't completely understand), I'd support it. Index: nsIParser.h =================================================================== RCS file: /cvsroot/mozilla/htmlparser/src/nsIParser.h,v retrieving revision 3.107 diff -u -r3.107 nsIParser.h --- nsIParser.h 2001/06/29 22:56:44 3.107 +++ nsIParser.h 2001/07/09 16:35:48 @@ -87,12 +87,12 @@ kCharsetFromWeakDocTypeDefault, kCharsetFromUserDefault , kCharsetFromDocTypeDefault, + kCharsetFromCache, kCharsetFromParentFrame, kCharsetFromBookmarks, kCharsetFromAutoDetection, kCharsetFromMetaTag, kCharsetFromByteOrderMark, - kCharsetFromCache, kCharsetFromHTTPHeader, kCharsetFromUserForced, kCharsetFromOtherComponent,
Strange. I got today's trunk build (20010709 debug) and meta charset seems working without nsIParser.h changes. I've asked ylong to test again.
Yes, I installed 07-09-09 trunk build, but couldn't reproduce this problem any more.
sr=jst
Fix checked into the trunk. It should make it into the branch when we move the fix for bug 81253 there.
Status: NEW → RESOLVED
Closed: 24 years ago
Resolution: --- → FIXED
Verified on 07-11-07 trunk build.
Whiteboard: verified on trunk
Mark it as verified per comments above.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.