Add unit tests for NTLM authentication. This requires...

Add unit tests for NTLM authentication.  This requires
overriding the functions that generate random bytes or get
the local host name, so that the generated NTLM messages are
reproducible.

R=eroman
BUG=6567, 6824
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=11488

[wtc, 2009-03-11 00:20:22]

Please review Patch Set 2.

Please be critical of the approach I use to mock the
GenerateRandom and GetHostName functions.

[wtc, 2009-03-11 01:52:26]

I added a second unit test, which tests the scenario
of entering a wrong password and then the right password.
Other than that, Patch Set 3 is the same as Patch Set 2.

[eroman, 2009-03-11 03:04:44]

lgtm

http://codereview.chromium.org/42052/diff/1008/13
File net/http/http_auth_handler_ntlm.cc (right):

http://codereview.chromium.org/42052/diff/1008/13#newcode448
Line 448: static HttpAuthHandlerNTLM::GenerateRandomProc generate_random_ =
optionally name |generate_random_proc_|.

Also, why not make this a static member of HttpAuthHandlerNTLM ? (since the
setters are).

http://codereview.chromium.org/42052/diff/1008/13#newcode450
Line 450: static HttpAuthHandlerNTLM::HostNameProc get_host_name_ = GetHostName;
ditto to above.

http://codereview.chromium.org/42052/diff/1008/13#newcode530
Line 530: host_len = strlen(host_buf);
what if GetHostName() returned a std::string instead?

http://codereview.chromium.org/42052/diff/1008/14
File net/http/http_auth_handler_ntlm.h (right):

http://codereview.chromium.org/42052/diff/1008/14#newcode21
Line 21: typedef void (*GenerateRandomProc)(void* output, size_t n);
might save some casts to define as uint8* (since each implementation ends up
doing the cast anyway).

http://codereview.chromium.org/42052/diff/1008/14#newcode23
Line 23: // A function that returns the local host name in the output buffer.
Returns
"returns" may not be the best wording here. Also worth mentioning
"null-terminated". See other comment about whether this can just return
std::string.

http://codereview.chromium.org/42052/diff/1008/14#newcode39
Line 39: static void SetGenerateRandomProc(GenerateRandomProc proc);
just thinking out loud: what if these were made into private methods? (Then you
would need a FRIEND_TEST() to set).

http://codereview.chromium.org/42052/diff/1008/12
File net/http/http_network_transaction_unittest.cc (right):

http://codereview.chromium.org/42052/diff/1008/12#newcode306
Line 306: for (size_t i = 0; i < n; i++)
nit: I think most of our other code does ++i.

http://codereview.chromium.org/42052/diff/1008/12#newcode317
Line 317: for (size_t i = 0; i < n; i++) {
same nit as above.

http://codereview.chromium.org/42052/diff/1008/12#newcode318
Line 318: buf[i] = bytes[current_byte++];
(This tripped me up when I first glanced at it; I then recognized it was postfix
form, therefore correct).

http://codereview.chromium.org/42052/diff/1008/12#newcode328
Line 328: name[0] = '\0';
(i guess no point checking if namelen == 0).