Memory Leak in RHTTPTransaction

[gammav]

I see some memory leaks in RHTTPTransaction.

Here is code snippet

iTrans = iSess.OpenTransactionL(uri, *iTransObs, aMethod);
..
...
iTrans.SubmitL();
CActiveScheduler::Start();

Immediately after iTrans.SubmitL() and in MHFRunL callback function, I can see that significant RAM is consumed. However, after completion of SubmitL, there is a portion of RAM which remains unrecovered.

Any information.

[wizard_hu_]

RHTTP... stuff uses string tables, that can be a cause for example.

[gammav]

RHTTP... stuff uses string tables, that can be a cause for example.

Posible, but user does not have any control over it, isn't?

Interestingly, it is always in multiple of 8192 and varies from 8192 to 32768

Is there something can be done solve it?

[wizard_hu_]

It may disappear when you Close the transaction. Do you get an ALLOC panic actually?

[gammav]

It may disappear when you Close the transaction. Do you get an ALLOC panic actually?

Wiz, transaction is closed properly in various conditons in MHFRunL. As you said, it does recover memory but not completely. Other ideas?

case THTTPEvent::ESucceeded:
{
aTransaction.Close();
RequestCompleted(aEvent.iStatus);
} break;
case THTTPEvent::EFailed:
{
aTransaction.Close();
RequestCompleted(aEvent.iStatus);
} break;

[wizard_hu_]

Do you get an ALLOC panic when you close your application in the emulator (having panic messages enabled, of course)? Transient memory allocation of system components is hard to influence.

[Paul.Todd]

There will be significant amounts of RAM consumed if the session is still open as this has all sorts of tables and the bearer open.

[gammav]

There will be significant amounts of RAM consumed if the session is still open as this has all sorts of tables and the bearer open.

Thanks Paul, session is properly closed in destructor

CMyClient::~CMyClient()
{
LOGRAMUSAGE();
iConnection.Close();
iSess.Close();
iFileServ.Close();
delete iReqBodySubmitBuffer;
delete iTransObs;
LOGRAMUSAGE();
}


Thanks Wiz, I do not get allocation panic.

The worst part is that these allocation does not get free even if I exit the application. My assumption is that if the leakage was in user application, it would have been free when application exited. Since it remains allocated, could that be in kernel?

[Paul.Todd]

If you have enabled extended panics as documented in the wiki and you still do not get a panic when exiting then all the memory is freed and all handles have been freed.

My guess is there is something wrong with your logging code.

[gammav]

My guess is there is something wrong with your logging code.

Hi Paul, logging code is fine and we use it for multiple projects. The problem is evident that after few hundre HTTP requests, phone starts showing the low memory symptoms and gives error like Memory full.

Similar discussion here though it may not apply completely

http://developer.symbian.com/forum/t...threadID=25888

[Paul.Todd]

That issue had the connection to the socket server returning an -4 error, your case is that memory applications within your application are failing.


It might be that your heap is very fragmented, but I almost certain that its a memory leak. You won't see the memory leaks on a real device as the devices have shipping ROM's, not debug ROMs.

Run in the emulator to see the leaks and use hooklogger (avaiable from Symbian's website to track down the allocations)

[Dmitry__]

I see some memory leaks in RHTTPTransaction.

Here is code snippet

iTrans = iSess.OpenTransactionL(uri, *iTransObs, aMethod);
..
...
iTrans.SubmitL();
CActiveScheduler::Start();

Immediately after iTrans.SubmitL() and in MHFRunL callback function, I can see that significant RAM is consumed. However, after completion of SubmitL, there is a portion of RAM which remains unrecovered.

Any information.

Hello gammav,
Have you solve an issue with memory leaking (worst case: OOM). If so, could you please share your opinion for solving the problem

With regards,
Dmitry__