Re: Interesting memory leaks ;-)
- From: Justin Schoeman <justin expertron co za>
- To: Justin Schoeman <justin expertron co za>
- Cc: "orbit-list gnome org" <orbit-list gnome org>
- Subject: Re: Interesting memory leaks ;-)
- Date: Mon, 05 Apr 2004 14:23:27 +0200
OK, some more details...
It seems that as long as the client has an open reference to the object,
then it maintains the connection (in this case IPv4) to the server. The
problems all seem to revolve around the server receiving a connection
close. When this happens, 1) memory is leaked, 2) if the reconnection
comes in too quickly, it may be stalled until retried.
I will look into these in more detail, but would appreciate any assistance!
TIA,
-justin
Justin Schoeman wrote:
> Hi all,
>
> My system is finally up and running, and with the long-term operation, I
> have noticed some memory leaks. I have put together some simple test
> applications, and run through some scenarios (which are detailed below).
> If anybody has some vague idea why these problems are occurring (before
> I start tearing orbit apart looking for it), I would appreciate it...
>
> [ All these scenarios describe the client side. The server stays the
> same, and is implemented based on the standard skeleton servers. ]
>
> Scenario #1:
> while(1) {
> import object1 from ior in file;
> call object1 method;
> release object1;
> }
>
> The server leaks memory.
>
> Scenario #2:
> import object1 from ior in file;
> while(1) {
> import object2 from ior in file;
> call object2 method;
> release object2;
> }
> release object1;
> No memory is leaked.
>
> Scenario #3:
> As in scenario #2, but the server is restarted while the client is in
> the loop.
> The server leaks memory.
>
> In all cases where memory leaks, the client application also stalls
> occasionally (for about 5 sec) before continuing. The server still
> responds to other client requests during the stall.
>
> It seems there is some problem on the server side with a client-side
> last_unref of an object?
>
> Any ideas what may be wrong, and where I should look?
>
> (Valgrind indicates that the leaked memory is in CORBA_ORB_init which
> can't be right?)
>
> Thanks,
> -justin
>
> _______________________________________________
> orbit-list mailing list
> orbit-list@gnome.org
> http://mail.gnome.org/mailman/listinfo/orbit-list
>
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]