Wireshark · Wireshark-dev: Re: [Wireshark-dev] RFD: Limiting scope of ep_ memory

[Jeff Morriss <jeff.morriss.ws@xxxxxxxxx>]

Evan Huus wrote:
> On Mon, Oct 22, 2012 at 12:21 PM, Anders Broman
> <anders.broman@xxxxxxxxxxxx> wrote:
> > - I think the whole ep/se memory idea is optimizing for execution speed and as a bonus fewer memory leaks
> >  Does this assumption still hold true e.g. ep_ is faster that g_malloc? (Changing back would be a nightmare I
> >  suppose). But it seems like having ep memory is geting quite complicated.
>
> I think reducing memory leaks is the primary goal, and execution speed
> was a bonus. I agree that it's getting overly complicated though, see
> my email from last week [1] for some thoughts on where I'd like to go
> long-term.

That's what my recollection says too: because even very simple dissector 
code like:

foo = g_malloc()
tvb_get_XXX()
g_free(foo);

Leaks memory when exceptions are thrown.

The addition of canaries to detect buffer overruns, etc. was another bonus.

Ah, yes, the wiki page:

http://wiki.wireshark.org/EMEMification

also lists fixing memory leaks as the primary reason.