The CertVerifierJob destructor should delete canceled requests....

The CertVerifierJob destructor should delete canceled requests.

Add a job to inflight_ only after the job's worker has started
successfully.

R=agl
BUG=63357, 67289
TEST=net_unittests --gtest_filter=CertVerifierTest.CancelRequestThenQuit
should not leak a CertVerifierRequest object under valgrind.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=72203

[wtc, 2010-12-18 01:12:19]

agl: your dnsrr_resolver.cc code doesn't have this memory
leak because the RRResolverJob destructor calls PostAll().
I didn't copy that code because I don't want the callbacks
to run when CertVerifier is deleted, but the problem is
that properly canceled requests are leaked.  (Post() not
only calls the callback, if not canceled, but also delete
itself.)

The inflight_ change was suggested by willchan in his
code review for CertVerifier.

[willchan no longer on Chromium, 2011-01-04 21:33:54]

LGTM

http://codereview.chromium.org/5973004/diff/4001/net/base/cert_verifier.cc
File net/base/cert_verifier.cc (right):

http://codereview.chromium.org/5973004/diff/4001/net/base/cert_verifier.cc#ne...
net/base/cert_verifier.cc:351: return ERR_FAILED;
Is this ERR_FAILED ok?  I don't know if we can choose a better error code.  This
reminds me, are we logging anything to the NetLog?  That's a good TODO item to
add if you haven't already.

http://codereview.chromium.org/5973004/diff/4001/net/base/cert_verifier_unitt...
File net/base/cert_verifier_unittest.cc (right):

http://codereview.chromium.org/5973004/diff/4001/net/base/cert_verifier_unitt...
net/base/cert_verifier_unittest.cc:262: scoped_ptr<CertVerifier> verifier(new
CertVerifier);
Nit: Does this need to be heap allocated?

[agl, 2011-01-06 14:03:00]

LGTM