Subj : Re: Still unsolved problem
To   : netscape.public.mozilla.jseng
From : David Bradley <dbradleyXSPAM@netscape.com>
Date : Wed Jun 04 2003 06:18 pm

Brendan Eich wrote:
> I'm not a Windows hacker, but I recall something about not mixing debug 
> and release mallocs.  Could that be happening?  Cc'ing dbradley, who 
> does know Windows.

Right, but I would think the memory in question here would have been allocated and freed from within the JS DLL. So things should be ok. If things were going wrong, I would expect more fatal problems than what he is seeing.

What is proxy_dns_resolve?
When you say error, is this a JS exception? How is it getting reported?
Also I see proxy_dnsResolve and proxy_dns_resove, I assume the latter is the implementation called by both proxy_isInNet and proxy_dnsResolve? If that's the case you might want to simplify your test and trying to just invoke the dnsResolve and see what happens.
I'd start by seeing what the value of host is in FindProxyForURL then check ipstr and ip in proxy_isInNet. You could dump this to a file fairly easily if you don't want to go to the trouble of building a release with symbols. (It's not that hard, though) I do the following:

In my .mozconfig
ac_add_options --disable-debug
ac_add_options --enable-optimize=-O1

Then I have MOZ_DEBUG=1 and MOZ_PROFILE=1, although I'm not sure MOZ_PROFILE is really needed.

But again you could just dump the values to a file fairly easily either using C++ streams or C streams and then inspect the file afterwards.

David

.