Closed Bug 49039 Opened 26 years ago Closed 25 years ago

national characters in a page intermittantly breaks rendering

Categories

(Core :: DOM: HTML Parser, defect, P3)

defect

Tracking

()

VERIFIED WORKSFORME

People

(Reporter: spam, Assigned: ftang)

Details

(Keywords: regression, Whiteboard: [nsbeta3] (data loss is VERY bad.))

first seen on linux M18 build 2000-081309 Noticed a query loaded from local disk broke some of the summary of bug 31919 "URLencode fails on URL with national ( æ ) character". As it turns out: ANY occurance of æ, ø or å in a html page will now break rendering of it. Tested "ñ" and "ö" - those too break pages. This bug doesn't break *saving*: The files are only trunkated when rendered. Tempted to set it as a blocker however: This bug effectively makes moz more or less completely useless for me. (Searches in bugzilla break, can't browser national wepages etc.) This bug is pretty new - days at most. Not sure of component - parser or localization or what.
regression keyw.
Keywords: regression
adding hint in sumary about not using moz to read this bug right now.
Summary: national characters in a page breaks loading → national characters in a page breaks loading (read this bug with NN, not moz!)
ahh this was fun: the error is intermittant. This page just loaded OK. Lowering severity but this should be looked in to.
Severity: critical → normal
Summary: national characters in a page breaks loading (read this bug with NN, not moz!) → national characters in a page intermittantly breaks rendering
wonder if i found a way to reproduce 100%: open an ftp site in one browser window. open a new browser window. Try to load this bug then. here, the first comment then ends this line: "URLencode fails on URL with national (
this is a pretty bad bug, as it means that a browser window launched from a window currently displaying ftp treeview of some site, isn't really a full fledged navigator window: Either it inherits "bad stuff" from treeview, or - for some reason - it isn't properly initiated.
OK, I can reproduce this with today's win32 and Mac builds. If I have a window open to an FTP site and from that window I launch a new window and try to laod http://bugzilla.mozilla.org/show_bug.cgi?id=49039 in this new window it is trunkated at the æ character.
OS: Linux → All
Hardware: PC → All
note: the source from view source is also truncated at that point. Setting to Platform/OS ALL
over to parser.
Assignee: asa → rickg
Component: Browser-General → Parser
QA Contact: doronr → janc
No real hard data yet, but this bug does not appear in viewer. That means 1) it's not parser; or 2) viewer's code path avoids the issue. Let's try mozilla now...
Status: NEW → ASSIGNED
Ok, it took a while to track it down, but here it is. The scanner gets the data correctly from netlib, and tries to append it (in the parser). The scanner calls the i18n library: res = mUnicodeDecoder->Convert(aBuffer, &srcLength, unichars, &unicharLength); which truncates the buffer after the 0xE6. I don't know why. Off to frank tang for more work.
Assignee: rickg → ftang
Status: ASSIGNED → NEW
Whiteboard: [nsbeta3] (data loss is VERY bad.)
Target Milestone: --- → M18
Please include reproduce procedure and /or test pages. With the current description, we have no way to reproduce / verify the issue. Please make some testing page for us. Thanks.
Status: NEW → ASSIGNED
ahh it's changed. With m18/linux/2000-082513 a browser-instance launched from another (while it displays an ftp site) does NOT break text in this bug anymore. Which was more or less how i discovered this bug. It seems to have passed away in silence.
i still see this in 2000091212 To reproduce: go to an ftp:// URL from that window: Open a new browser window in the new window: go to http://www.dagbladet.no All national characters (ז,ר,ו) will now display as "?" (There is a cache bug also, so quitting mozilla and then starting again and going to dagbladet.no will now show "?" instead of זרו even if cache is set to refresh once per session. To verify that the site actually IS coded right, one has to go to another site, delete disk and memory cache from prefs, close mozilla, and then restart and go there again. Characters should now display זרו) I think bug 52255 is a dup of this one.
Oops - forgot the vital piece of info here: The bug no longer breaks all the rest of the page, but it's still data-loss that it looses the ability to interpret character-sets. Any language with more national characters than norwegian can get very unreadable very quick as long as this bug isn't fixed.
I cannot reproduce this. Mark it as workforme.
Status: ASSIGNED → RESOLVED
Closed: 25 years ago
Resolution: --- → WORKSFORME
updated qa contact.
QA Contact: janc → bsharma
Verified on build: 2001-05-07-06-Trunk platform: WinNT
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.