208127 - nsXMLHttpRequest() Crash after 501 calls to PushThreadEventQueue

Status: RESOLVED DUPLICATE of bug 206947

(Core :: XPCOM, defect)

Description

[Narayanan Govindaraju]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.1) Gecko/20020826
Build Identifier: Mozilla/5.0 (X11; U;Linux i686; en-US; rv:1.2.1) Gecko/20030404

<script language="JavaScript" type="text/JavaScript">
var xmlbase;
xmlbase = new XMLHttpRequest();
xmlbase.open("GET","xyz.xml",false);
xmlbase.send(null);
xmlbase=null;
delete xmlbase;
</script>

Memory is not released evenafter unloading this page and loading new page.


Reproducible: Always

Steps to Reproduce:
1. create a web page with XMLHttpRequest() as follows
var xmlbase;
xmlbase = new XMLHttpRequest();
xmlbase.open("GET","xyz.xml",false);
xmlbase.send(null);
xmlbase=null;
delete xmlbase;
2. set this page to auto refresh to simple page(image only) in 10 sec. 
3. That simple page auto refresh back to this page in 10 sec
(that is two page slide show one with XMLHttpRequest() and another simple page)
4. load the page 1 in Mozilla

Actual Results:  
Watch out the memory usage. Every time it loads the page with XMLHttpRequest()
memory usage piles up.

Expected Results:  
Whenever it unloads a page memory should be released

Latest release also have this problem

Comment 1

[Narayanan Govindaraju]

Attachment: Page 1 with XMLHttpRequest

Comment 2

[Narayanan Govindaraju]

Attachment: Page 2 (Simple Page)

Comment 3

[Narayanan Govindaraju]

Attachment: XML File

Comment 4

[Narayanan Govindaraju]

Attachment: XML File

Comment 5

[Heikki Toivonen (remove -bugzilla when emailing directly)]

I don't see XMLHttpRequest leaking. After waiting for 3 reloads of each of the
document I shut Mozilla down and these are the reported leaks:

== BloatView: ALL (cumulative) LEAK STATISTICS

    
|<----------------Class--------------->|<-----Bytes------>|<----------------Objects---------------->|<--------------References-------------->|
                                              Per-Inst   Leaked    Total     
Rem      Mean       StdDev     Total      Rem      Mean       StdDev
   0 TOTAL                                          47      336   306057       
6 (  757.68 +/-  1148.19)  2381086        3 ( 1194.46 +/-  2026.84)
nsTraceRefcnt::DumpStatistics: 600 entries
  99 XPCNativeScriptableShared                      76       76      547       
1 (   20.85 +/-     6.42)        0        0 (    0.00 +/-     0.00)
 102 XPCWrappedNative                               52       52     1864       
1 (  492.16 +/-   266.31)    36711        1 (  826.01 +/-   341.96)
 331 nsHashtable                                    44       88      745       
2 (  313.97 +/-   175.47)        0        0 (    0.00 +/-     0.00)
 382 nsMemoryImpl                                   72       72        2       
1 (    1.33 +/-     0.58)       36        1 (    5.97 +/-     1.51)
 565 nsXPCComponents                                48       48       26       
1 (    7.65 +/-     3.15)      935        1 (   26.60 +/-     5.98)

Until you can show XMLHttpRequest really is leaking I am marking this worksforme.

Comment 6

[Greg K.]

*** Bug 208138 has been marked as a duplicate of this bug. ***

Comment 7

[Narayanan Govindaraju]

It is not the problem of releasing the memory during browser shutdown. It is
releasing the memory when moving to new page. Let these 2 page run for some time
and watch the memory free. Whenever XMLHttprequest page is loaded memory usage
getting piled up. 

If it not because of XMLHttprequest, can we assigned to JavaScript engine or
other component, which is not releasing the memory when it goes to next page.

Comment 8

[Pete Collins (MDG)]

This is a threading issue. Changing component to XPCOM.

Comment 9

[Pete Collins (MDG)]

Attachment: updated page1.html test

Comment 10

[Pete Collins (MDG)]

Attachment: updated page2.html

Comment 11

[Pete Collins (MDG)]

Attachment: a plain text file the test pages read

Comment 12

[Pete Collins (MDG)]

Attachment: Test patch to count calls to PushThreadEventQueue

Comment 13

[Pete Collins (MDG)]

Steps to reproduce crash:

- Place attachment (127746), (127747), (127748) on a web server.
- Patch nsXMLHttpRequest.cpp w/ attachment (127749)
- recompile extensions/xmlextras module
- Load page1.html into browser
- let refresh run until it crashes.
- Crash will happen after exactly 501 calls to PushThreadEventQueue 

http://lxr.mozilla.org/seamonkey/source/extensions/xmlextras/base/src/nsXMLHttpRequest.cpp#1282

I'm looking into this now. It appears that 501 might be a magic recursion
blocking number?

Any comments?

Comment 14

[Andrew Schultz]

this looks like bug 206947

Comment 15

[Pete Collins (MDG)]

*** This bug has been marked as a duplicate of 206947 ***