Add a unit test for handling SSL certificate errors....

Add a unit test for handling SSL certificate errors.

Add CreateSSLClientSocket method to eliminate duplicate code in tests.

R=eroman
BUG=none
TEST=new unit test

[wtc, 2009-08-14 22:39:27]

This is work in progress.  I need to wait for Chris Hawk to fix
http://crbug.com/14733 before I can continue.

I added the unit test to ssl_client_socket_unittest.cc.  Unfortunately
requires me to copy some code from http_network_transaction.cc.
Ideally I should add the tests to http_network_transaction_unittest.cc,
but all the tests in that file seem to use mock sockets.  I need to
figure out how to add tests to http_network_transaction_unittest.cc
that use real SSL sockets.

[eroman, 2009-08-15 02:04:28]

http://codereview.chromium.org/170016/diff/1/2
File net/socket/ssl_client_socket_unittest.cc (right):

http://codereview.chromium.org/170016/diff/1/2#newcode57
Line 57: net::ClientSocket *transport = new net::TCPClientSocket(addr);
style-nit: move * to the left.

http://codereview.chromium.org/170016/diff/1/2#newcode168
Line 168: EXPECT_EQ(net::CERT_STATUS_AUTHORITY_INVALID,
style-nit: why not just EXPECT_TRUE(ssl_info.cert_status &
net::CERT_STATUS_AUTHORITY_INVALID) ? or can you do
EXPECT_EQ(net::CERT_STATUS_AUTHORITY_INVALID, or ssl_info.cert_status) ?

http://codereview.chromium.org/170016/diff/1/2#newcode176
Line 176: ///////////////////////
style-nit: i haven't really seen the //// style in other places. I am not
opposed to it, just wandering if there is precedent.

http://codereview.chromium.org/170016/diff/1/2#newcode219
Line 219: CreateSSLClientSocket(addr, kDefaultSSLConfig));
Is this intentional? This used to use server_.kMismatchedHostName, but now it
uses server_.kHostName