89643 - m92 build is busted in non ASCII pages

Status: VERIFIED FIXED

(Core :: Internationalization, defect, P1)

Description

[Frank Tang]

I found this yesterday while I looking with marina 
It seems the m92 is also busted with different sympton.
1. When I visit http://home.netscape.com/ja some of the image never finish 
loading. If I stop then reload, it looks fine. I wonder is that related to 
82244...
2. When I visit http://home.netscape.com/ko , zh/tw or zn/cn it never finish the 
loading

Comment 1

[Frank Tang]

talk to ylong, we can start to reproduce this from 070503 m92 build. we cannot 
reproduce on 070503 m92 build.

Comment 2

[Frank Tang]

The problem probably not from my check in of 82244 since I land it 6/29 and the 
0703 build is still good. 

Comment 3

[Frank Tang]

adding any one who check in into m92 during 0703 3:00 to 0705 3:00 to the cc 
list- 
waterson, dougt, jgaunt, jst

Comment 4

[Yuying Long]

This problem is existing on all platform.

Btw, it this related to bug 81253? I checked some other CJK pages, if I click 
stop(like netscape.com/zh/cn), and reload it, then will loading OK.

Comment 5

[Frank Tang]

remeber, before you try to reproduce this problem, clear your cache first.

Comment 6

[Frank Tang]

I am really lucky, find out the problem at once single try.
I back out dougt's chagne in  mozilla/ netwerk/ protocol/ http/ 
src/nsHttpChannel.cpp rev 1.31.2.3 checked in 07/04/2001 08:10 and the problem 
go away.

Comment 7

[Frank Tang]

reassign to dougt

Comment 8

[Frank Tang]

With this bug, I don't think tester can pass smoke test if their cache is clear 
and don't try reload.

Comment 9

[Frank Tang]

I think this should be a PDT+ bug. 

Comment 10

[Frank Tang]

dougt' check in for nsHttpChannel.cpp rev 1.31.2.3  is for bug 82418

Comment 11

[Doug Turner (:dougt)]

sorry about that.  Backing out.

Checking in nsHttpChannel.cpp;
/cvsroot/mozilla/netwerk/protocol/http/src/nsHttpChannel.cpp,v  ?--  
nsHttpChannel.cp
new revision: 1.31.2.4; previous revision: 1.31.2.3
done

Marking as fixed.  Frank, can you please verify?

Comment 12

[Doug Turner (:dougt)]

*** Bug 89472 has been marked as a duplicate of this bug. ***

Comment 13

[selmer (gone)]

Doug, I assume that you checked in on the branch and the PDT+ is redundant :-)

Comment 14

[Frank Tang]

reopen. dougt, you didn't back it out.
I see some other problems.
Here is the difference between before your check in and after your backout. 
Which is different. I expect it contains zero diff.


Z:\mozilla\netwerk\protocol\http\src>cvs diff -r 1.31.2.2 -r 1.31.2.4 -u nsHttp
hannel.cpp
Index: nsHttpChannel.cpp
===================================================================
RCS file: /cvsroot/mozilla/netwerk/protocol/http/src/nsHttpChannel.cpp,v
retrieving revision 1.31.2.2
retrieving revision 1.31.2.4
diff -u -r1.31.2.2 -r1.31.2.4
--- nsHttpChannel.cpp   2001/06/27 14:48:51     1.31.2.2
+++ nsHttpChannel.cpp   2001/07/06 19:41:49     1.31.2.4
@@ -967,12 +967,9 @@
     rv = mCacheTransport->OpenOutputStream(0, PRUint32(-1), 0, getter_AddRefs(
ut));
     if (NS_FAILED(rv)) return rv;

-    // XXX disk cache does not support overlapped i/o yet
-#if 0
     // Mark entry valid inorder to allow simultaneous reading...
     rv = mCacheEntry->MarkValid();
     if (NS_FAILED(rv)) return rv;
-#endif

     nsCOMPtr<nsIStreamListenerTee> tee =
         do_CreateInstance(kStreamListenerTeeCID, &rv);
@@ -2104,8 +2101,11 @@
         mTransaction = nsnull;
     }

+    // we don't support overlapped i/o (bug 82418)
+#if 0
     if (mCacheEntry && NS_SUCCEEDED(status))
         mCacheEntry->MarkValid();
+#endif

     if (mListener) {
         mListener->OnStopRequest(this, mListenerContext, status);

Comment 15

[Frank Tang]

so.... rev 1.31.2.2 have no problem
rev1.31.2.3 cause image never finish loading in http://home.netscape.com/ja and 
http://home.netscape.com/ko never finish loading
rev1.31.2.4 cause http://home.netscape.com/ja never show up.

Please back it out back to 1.31.2.2. your check in in1.31.2.4 is not identical 
as 1.31.2.2

Comment 16

[Frank Tang]

ok. talk to dougt. he definitely need more sleep. I think he back out the code 
in a wrong way some how add back #if into line 2000+ instaead of 900+
reback out to r 1.31.2.2 for him :)
mark it fixed

Comment 17

[Yuying Long]

This problem has been resolved on recently branch and trunk build.