104002 - Leaking 600 bytes in nsIOService::Create switching between 2 newsgroups.

Status: VERIFIED WORKSFORME

(Core :: Networking, defect)

Description

[[e-mail] (gone - use [e-mail] instead)]

Current trunk, Windows 2000.

Steps to Reproduce:

1.  Have 2 newsgroups set up, I had netscape.public.test and netscape.test.
2.  Switch between them (no need to click on messages).

[W] MLK: Memory leak of 600 bytes from 2 blocks allocated in nsIOService::Create
(nsISupports *,nsID const&,void * *)
        Distribution of leaked blocks
               400 bytes from 1 block of 400 bytes (0x042f3328) 
               200 bytes from 1 block of 200 bytes (0x04301790) 
        Allocation location
            PR_Malloc+0x2c       [.\../../../../pr/src/malloc/prmem.c:50 
ip=0x30018c5c]
            nsIOService::Create(nsISupports *,nsID const&,void * *)+0x12b 
[c:\moz_src\mozilla\netwerk\base\src\nsIOService.cpp:266 ip=0x047e078e]
            nsGenericFactory::CreateInstance(nsISupports *,nsID const&,void * *)
+0x50 [c:\moz_src\mozilla\xpcom\components\nsGenericFactory.cpp:74 
ip=0x100a734f]
            nsComponentManagerImpl::GetService(nsID const&,nsID 
const&,nsISupports * *,nsIShutdownListener *)+0x3f8 
[c:\moz_src\mozilla\xpcom\components\nsComponentManager.cpp:1763 ip=0x100af97b]
            nsServiceManager::GetService(nsID const&,nsID const&,nsISupports * 
*,nsIShutdownListener *)+0x95 
[c:\moz_src\mozilla\xpcom\components\nsComponentManager.cpp:3351 ip=0x100b636f]
            nsGetServiceByCID::()(nsID const&,void * *)const+0xa4 
[c:\moz_src\mozilla\xpcom\components\nsComponentManager.cpp:225 ip=0x100a90f3]
            nsCOMPtr_base::assign_from_helper(nsCOMPtr_helper const&,nsID 
const&)+0x56 [c:\moz_src\mozilla\xpcom\glue\nsCOMPtr.cpp:80 ip=0x10104b29]
            NS_NewURI(nsIURI * *,char const*,nsIURI *,nsIIOService *)+0x97 
[.\..\..\..\dist/include/necko\nsNetUtil.h:92 ip=0x0476542f]
            RDFXMLDataSourceImpl::Init(char const*)+0xc6 
[c:\moz_src\mozilla\rdf\base\src\nsRDFXMLDataSource.cpp:646 ip=0x04765163]
            nsChromeRegistry::AddToCompositeDataSource(int)+0x83f 
[c:\moz_src\mozilla\rdf\chrome\src\nsChromeRegistry.cpp:2699 ip=0x0465a427]
            nsChromeRegistry::LoadInstallDataSource(void)+0x9a 
[c:\moz_src\mozilla\rdf\chrome\src\nsChromeRegistry.cpp:2882 ip=0x0465b9e0]
            nsChromeRegistry::Init(void)+0xbff 
[c:\moz_src\mozilla\rdf\chrome\src\nsChromeRegistry.cpp:362 ip=0x046492b2]
            nsChromeRegistryConstructor+0xeb 
[c:\moz_src\mozilla\rdf\chrome\build\nsChromeFactory.cpp:51 ip=0x0464323a]
            nsGenericFactory::CreateInstance(nsISupports *,nsID const&,void * *)
+0x50 [c:\moz_src\mozilla\xpcom\components\nsGenericFactory.cpp:74 
ip=0x100a734f]
            nsServiceManager::GetService(nsID const&,nsID const&,nsISupports * 
*,nsIShutdownListener *)+0x95 
[c:\moz_src\mozilla\xpcom\components\nsComponentManager.cpp:3351 ip=0x100b636f]
            nsGetServiceByCID::()(nsID const&,void * *)const+0xa4 
[c:\moz_src\mozilla\xpcom\components\nsComponentManager.cpp:225 ip=0x100a90f3]
            nsCOMPtr_base::assign_from_helper(nsCOMPtr_helper const&,nsID 
const&)+0x56 [c:\moz_src\mozilla\xpcom\glue\nsCOMPtr.cpp:80 ip=0x10104b29]
            InstallGlobalLocale+0xb1 
[c:\moz_src\mozilla\xpfe\bootstrap\nsAppRunner.cpp:869 ip=0x004079ff]
            main1+0x855          
[c:\moz_src\mozilla\xpfe\bootstrap\nsAppRunner.cpp:1178 ip=0x0040426f]
            main+0x262           
[c:\moz_src\mozilla\xpfe\bootstrap\nsAppRunner.cpp:1621 ip=0x004036df]

Comment 1

[Chase Tingley]

This is odd.  From this stack it looks like a new nsIOService is being created
(and leaked) every time we open one of the datasources.  That doesn't sound right.

Comment 2

[Vinay Badami]

I was not able to reproduce this leak either. 
Can u retry this using a debug build on the tip ?
I have tried
1. Using purify
2. Putting breakpoints in the debugger in the functions that are supposed to
leak. They do not get repeatedly called at all.

Comment 3

[[e-mail] (gone - use [e-mail] instead)]

marking wfm.

This could've been fixed by a number of things, one of which *might* have been 
bug 110850 (or any of Darin's recent URL/netwerk fixes).

Comment 4

[[e-mail] (gone - use [e-mail] instead)]

Ben, as reporter, it's safe to verify this for me.

Comment 5

[[e-mail] (gone - use [e-mail] instead)]

verified