71740 - Transfer-Encoding: articles frequently do not display or crash; M09 crash [@ MSVCRT.dll - nsWriteToRawBuffer ]

Status: VERIFIED FIXED

(Core :: Networking: HTTP, defect)

Description

[Ninoschka Baca]

Build 2001-03-06-05: NT4

Overview: When I get a stock quote and select the "News" link to see the latest
articles on a company, very often nothing displays.

Steps to reproduce:
1. Go to www.netscape.com
2. In the Business section where it states "Stock Quotes by CITITRADE", 
enter a stock symbol such as "dcnt" (or webm), select Quote button.
3. When the next page appears select the "News" link and it shows a 
variety of links to articles
4. Select a link to an article

Actual Results: Very often the article never appears. If the same steps 
are followed in 4.7 then it works. 
Note: this is not a Sidebar specific problem.

Expected Results: The article should display just as it does in 4.x.

Comment 1

[chris hofmann]

played with this a while on win32 laptop with dialup connection.
1) don't see Stock Quotes by CITITRADE 
   - has it changed to Salomon Smith Barney?
2) smith barney stock searches seem to show up ok...

is anyone else seeing this problem?

Comment 2

[lchiang]

I reproduced this with the following:
winNT build 2001-04-18-09 trunk.

1.  Either go to www.netscape.com or go to Stocks panel
2.  On www.netscape.com, enter webm in the quote field (business section) and
get the quote (or enter this in the stock sidebar)
3.  Stock quote appears in the browser window
4.  Click News link
5.  Click the News article entitled "WebMethods, Ford to announce deal" dated
4-13-2001.
6.  Blank page loads, no article appears.

I'm wondering if specific to certain conditions?  I didn't have a problem when I
tried an article from the dcnt quote/news page.

Comment 3

[lchiang]

cc: chofmann so he can see my response.

Comment 4

[chris hofmann]

ok, I see, as the bind man said as he picked up his hammer and saw...

here is the url that gets directly to the problem...

http://werpfarm.netscape.com/ns/pf/news/ticker/story.tmpl?url=/news/feeds/pf/tic
ker_search_reuters_spo/defaultCA=0D340510_7&display=/ns/pf/news/ticker/story_det
ail&symbol=webm

we had some problems loading content on urls that had ","
in them a few days ago.  I wonder if this is related...

I think it was darin that was studyin'/fixin' that bug.

Comment 5

[chris hofmann]

darin, can you have a look to see if my theory is correct?

Comment 6

[Darin Fisher]

benc: do you think you could generate a packet trace and/or HTTP log for this
page to see how much we're actually downloading?  thx!

Comment 7

[lchiang]

Darin, I have a bug filed by Sol where clicking on any stock news articles 
served by werpfarm.netscape.com results in a crash instead of a blank page. 
(Same NT 4.0 platform as reported here.)  Can you look at the stack trace and 
let me know if you believe it to be the same bug as this or a different bug?  I 
think it may be similar since all steps are the same and it's the results that 
differ - crash or blank page.  Thank you.

nsWriteToRawBuffer 
[d:\builds\seamonkey\mozilla\xpcom\io\nsPipe2.cpp, line 521]
nsPipe::nsPipeInputStream::ReadSegments
[d:\builds\seamonkey\mozilla\xpcom\io\nsPipe2.cpp, line 412]
nsPipe::nsPipeInputStream::Read                                
[d:\builds\seamonkey\mozilla\xpcom\io\nsPipe2.cpp, line 529]
nsHTTPChunkConv::OnDataAvailable        
[d:\builds\seamonkey\mozilla\netwerk\streamconv\converters\nsHTTPChunkConv.cpp, 
line 340]
nsHTTPServerListener::OnDataAvailable 
[d:\builds\seamonkey\mozilla\netwerk\protocol\http\src\nsHTTPResponseListener.cp
p, line 540]
nsOnDataAvailableEvent::HandleEvent                    
[d:\builds\seamonkey\mozilla\netwerk\base\src\nsStreamListenerProxy.cpp, line 
183]
PL_HandleEvent                                
[d:\builds\seamonkey\mozilla\xpcom\threads\plevent.c, line 589]
PL_ProcessPendingEvents                                 
[d:\builds\seamonkey\mozilla\xpcom\threads\plevent.c, line 522]
_md_EventReceiverProc                                 
[d:\builds\seamonkey\mozilla\xpcom\threads\plevent.c, line 1070]

Comment 8

[Darin Fisher]

yeah... i could believe that there is something screwy going on with the 
chunked decoder.

Comment 9

[lchiang]

thank you, Darin. I will use this bug to monitor the crash as well then.

Comment 10

[[e-mail]]

I'll try to reproduce this and if I can, I'll try to trace it. If I cannot, I'll
try to get a trace of Lisa's system.

Comment 11

[Jan Carpenter]

This bug is a topcrasher, added topcrash keyword.  
Added M09 crash [@ MSVCRT.dll - nsWriteToRawBuffer ] to summary for tracking.  

Here is a recent stack trace:

Incident ID 29714570 
MSVCRT.dll + 0x1637 (0x78001637) 
nsWriteToRawBuffer [d:\builds\seamonkey\mozilla\xpcom\io\nsPipe2.cpp, line 521] 
nsPipe::nsPipeInputStream::ReadSegments 
[d:\builds\seamonkey\mozilla\xpcom\io\nsPipe2.cpp, line 412] 
nsPipe::nsPipeInputStream::Read 
[d:\builds\seamonkey\mozilla\xpcom\io\nsPipe2.cpp, line 529] 
nsHTTPChunkConv::OnDataAvailable 
[d:\builds\seamonkey\mozilla\netwerk\streamconv\converters\nsHTTPChunkConv.cpp, 
line 340] 
nsHTTPServerListener::OnDataAvailable 
[d:\builds\seamonkey\mozilla\netwerk\protocol\http\src\nsHTTPResponseListener.cp
p, line 540] 
nsOnDataAvailableEvent::HandleEvent 
[d:\builds\seamonkey\mozilla\netwerk\base\src\nsStreamListenerProxy.cpp, line 
183] 
PL_HandleEvent [d:\builds\seamonkey\mozilla\xpcom\threads\plevent.c, line 589] 
PL_ProcessPendingEvents [d:\builds\seamonkey\mozilla\xpcom\threads\plevent.c, 
line 522] 
_md_EventReceiverProc [d:\builds\seamonkey\mozilla\xpcom\threads\plevent.c, line 
1070] 

Comment 12

[Gagan]

may be invalidated by bug 76866. Tom/Benc could you guys verify this after 76866
lands? Thanks.

Comment 13

[lchiang]

gagan - if you can provide an experimental build for me, I can try this for you
before any landings occur.  If not fixed by the landing, then we'll know sooner
:-)  I've never seen the crash, but I get the problem with not displaying.

Comment 14

[Darin Fisher]

i played around a bit on these sites with a build post http branch landing, and
things seem ok.  marking WORKSFORME (which may be a bit optimistic).

Comment 15

[Ninoschka Baca]

Build 2001-05-11-04: WinMe. I still see the problem even when trying to view
articles on AOL.

Comment 16

[lchiang]

Darin - does "with a build post http branch landing" mean a build after 5-11? 
When did your branch get carpooled in? Thanks.

Comment 17

[Phil Peterson]

Still seeing MyNetscape pages not loading using the 5/16 morning build. Not
seeing crashes.

Comment 18

[Phil Peterson]

reopening

Comment 19

[Phil Peterson]

*** Bug 77853 has been marked as a duplicate of this bug. ***

Comment 20

[lchiang]

This is still a problem on Win32 2001-05-15 builds. If your changes fixed 
this, please mark fixed once you have your changes checked into the trunk.

Comment 21

[Darin Fisher]

my changes went in on 5/11, so if we are still seeing this then no my changes
didn't solve the problem.  but, it does sound like my changed helped make this
not a crasher.  am i correct about that?

Comment 22

[Jan Carpenter]

I don't see a talkback report of this after 5/12, so I'm guessing it's not a 
crasher anymore.

Comment 23

[Darin Fisher]

ok, i understand the problem here.  the server is sending a "Transfer-Encoding:
identity" header, which we are not correctly handling.  The spec actually says
that if a Transfer-Encoding header is sent, that it must contain the value
"chunked", which this one does not!

Comment 24

[Darin Fisher]

but we can make a work around for this very easily.  patch in hand....

Comment 25

[Darin Fisher]

Attachment: fixes the problem

Comment 26

[Darin Fisher]

Attachment: the real patch (w/o other junk that was sitting in my tree)

Comment 27

[Gagan]

change PL_strstr to PL_strncasestr since "chunked" is always 7 and with that
r=gagan

Comment 28

[Darin Fisher]

Attachment: much improved patch (thanks gagan!)

Comment 29

[Darin Fisher]

Attachment: let me try that again!

Comment 30

[Gagan]

r=gagan

Comment 31

[Doug Turner (:dougt)]

sr

Comment 32

[Darin Fisher]

fix checked in

Comment 33

[Darin Fisher]

*** Bug 59464 has been marked as a duplicate of this bug. ***

Comment 34

[Andreas M. "Clarence" Schneider]

Verified.
BTW:
1) We crash if a server says "Transfer-Encoding: chunked" and that's not true.
   I'll investigate that and file a new bug.
2) RFC 2616 is not clear about "identity" without "chunked". It referrs to a
   section missing in the RFC (3.6.2). If the connection is terminated by
   closing the connection "chunked" is never required.

Comment 35

[lchiang]

[e-mail] - which platform and build did you verify on?  Thanks.

Comment 36

[Andreas M. "Clarence" Schneider]

lchiang: It was a yesterday's CVS build on Linux. Also confirmed with
2001-05-23-20 on Win NT (for the crash a talkback id is TB30847740Q).

Comment 37

[Andreas M. "Clarence" Schneider]

My comment on the crash was wrong. By mistake I sent Content-Encoding, not
Transfer-Encoding. Nevertheless I filed bug 82577.