So I double checked master with the heap checker and it seems like this was
a false alarm. It looks like all the leaks are in the machine builder and
the stats builder which aren't an issue.



On Fri, Oct 28, 2011 at 5:47 PM, avadh patel <[email protected]> wrote:

>
>
> On Fri, Oct 28, 2011 at 8:38 AM, DRAM Ninjas <[email protected]> wrote:
>
>> Greetings,
>>
>> We've been memory leak hunting recently and I'm not sure if this is an
>> actual leak, but I think it's something to keep in mind
>>
>> In particular I'm referring to the stringbuf that holds the history for a
>> memory request. The way this currently works is that both
>> MemoryRequest::init() functions check if there's a history pointer -- and if
>> there is, they delete it.
>>
>> Now, I haven't gone through and checked all the code paths, but this
>> requires all callers to do get_free_request() followed by init(). Now, you
>> could imagine some code failing to call init(), maybe by like setting the
>> fields manually directly or by calling reset() -- none of these alternate
>> methods will free the memory allocated for the history buffer. I've been too
>> lazy to check all these code paths, but I'm running the gperf heap checker
>> on it now to double check.
>>
>> Now, I think the way it's done this way is to preserve the history buffer
>> for free requests in case something goes wrong, but I think there has to be
>> a better way to do this. Perhaps implementing:
>>
>> get_free_request(MemoryRequest *m); // get a free request and call init(m)
>> to populate it with data
>> and
>> get_free_request(); // get a free request and call init() for the user to
>> get a blank one
>>
>
> I agree with your suggestions.  Its better to have code that does cleanup
> internally.
> I'll keep this in a TODO list for next release.
>
> Thanks,
> Avadh
>
>
>>
>>
> Anyways, I thought I'd point it out even though it's not a big deal.
>>
>>
>>
>> _______________________________________________
>> http://www.marss86.org
>> Marss86-Devel mailing list
>> [email protected]
>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>>
>>
>
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel

Reply via email to