13742 - MLK: 4 bytes leaked- www.visa.com - nsDNSService::Lookup()

Status: VERIFIED WONTFIX

(Core :: Networking, defect, P3)

Description

[Bruce Mitchener]

Solaris 2.6, gcc 2.7.2.3.  Build from September 13, 1999.  Opened viewer, loaded
initial page, went to www.visa.com, selected the 'back to school' specials link,
quit.

MLK: 4 bytes leaked at 0x33b988
  * This memory was allocated from:
        malloc         [rtlib.o]
        strdup         [libc.so.1]
        __rpc_getconfip [rpc_generic.c]
        gethostbyname_r [gethostbyname_r.c]
        PR_GetHostByName [prnetdb.c:309]
        nsDNSService::Lookup(nsISupports*,const
char*,nsIDNSListener*,nsIRequest**) [nsDnsService.cpp:172]
        nsSocketTransport::doResolveHost() [nsSocketTransport.cpp:555]
        nsSocketTransport::Process(short) [nsSocketTransport.cpp:450]
        nsSocketTransportService::ProcessWorkQ()
[nsSocketTransportService.cpp:243]
        nsSocketTransportService::Run() [nsSocketTransportService.cpp:448]
        nsThread::Main(void*) [nsThread.cpp:102]
        _pt_root       [ptthread.c:158]
        _thread_start  [libthread.so.1]

Comment 1

[Warren Harris]

This can wait until after m12.

Comment 2

[leger]

Bulk move of all Necko (to be deleted component) bugs to new Networking

component.

Comment 3

[Warren Harris]

Post beta.

Comment 4

[gordon]

It looks like gethostbyname leaks 4 bytes on Solaris.  We need to figure out if
this leak occurs once per invocation, or once for the life of the app.

Comment 5

[gordon]

I think this is more an issue for NSPR.  Wan-Teh, can you see this gets to the 
right place?

Comment 6

[Wan-Teh Chang]

I know about this memory leak in Solaris's gethostbyname_r().
We can send a bug report to Sun.  We can call gethostbyname()
in other short-lived processes (e.g., dns-helper).  I would
not worry about it.  

Comment 7

[Wan-Teh Chang]

I asked a Sun engineer to file a Solaris bug report
for me.  I suspect that this is a known bug.

I do not plan to work around this bug in NSPR.
Reassigned bug to gordon.

Comment 8

[gordon]

We suspect that the Solaris gethostbyname_r() is allocating some thread specific 
storage, and that it doesn't leak everytime it is called.  Sun has been made 
aware of the problem, so I'm marking this WONTFIX.

Comment 9

[Bruce Mitchener]

Thanks.  No objection from me.