Adds support for the credentials frame.

Adds support for the credentials frame.

Need to allow spdy sessions to see if the underlying socket is required
OBCs. If so, and if pooling a connection, then we need to generate a
CREDENTIALS frame for this origin.

Need to add support for the RST_FRAME status code of INVALID CREDENTIALS

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=120923

[wtc, 2012-01-28 00:41:20]

Review comments on Patch Set 1:

I only had time to review ssl_client_socket.*,
ssl_client_socket_nss.*, and spdy_credential_state.*.
Sorry.

It seems that the CL is work in progress.  Here are the
comments on the code I have reviewed so far.

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.cc
File net/socket/ssl_client_socket.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.cc...
net/socket/ssl_client_socket.cc:15: was_origin_bound_cert_sent_(false) {

origin_bound_cert_type_ is not initialized.

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.h
File net/socket/ssl_client_socket.h (right):

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.h#...
net/socket/ssl_client_socket.h:173: SSLClientCertType origin_bound_cert_type_;

It seems that origin_bound_cert_type_ can replace the
was_origin_bound_cert_sent_ field.

We can infer was_origin_bound_cert_sent_ by testing
origin_bound_cert_type_ != CLIENT_CERT_INVALID_TYPE.

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket_ns...
File net/socket/ssl_client_socket_nss.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket_ns...
net/socket/ssl_client_socket_nss.cc:2725: }

SSLClientSocketNSS needs to set origin_bound_cert_type_
(which requires origin_bound_cert_type_ to be a protected
member) or call set_origin_bound_cert_type().

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state.cc
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:17: last_added_(-1) {}

Since npos and last_added_ are size_t, which is unsigned,
it may be problematic to set them to -1.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:29: }

In src/net, we usually omit curly braces in one-liner if
statements.  Please follow this style in your code.

NOTE: Darin also made me use curly braces if there is an
"else", even when both the "if" branch and "else" branch
are one line.  For example,
  if (condition) {
    ++foo;
  } else {
    foo = 0;
  }

This could be Darin's personal preference, so I recommend
it but won't insist on it.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:46: if (slots_[i].Equals(origin)) {

Nit: It should be possible to replace this for loop with some
STL technique.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:50: return -1;

Change -1 to npos.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:58: }

I believe slots_.resize(size) can also be used to truncate.
Could you find out if that's the case?

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state.h
File net/spdy/spdy_credential_state.h (right):

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:18: // make sure that we store the OBC
properly.

This comment sounds like a TODO comment or a reminder for
yourself.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:21: class NET_EXPORT_PRIVATE
SpdyCredentialState {

Nit: it seems that this class should be named SpdyCredentialStore
because you refer to "the store" in the comments below.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:23: static size_t default_num_slots;

The (static) data member should be listed after the methods.

The data member name should end with an underscore '_'.

This data member should be 'const'.  So it may need to
follow the naming convention of constants.  Please check
the Style Guide.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:27: // Changes the size of the store.  If the
new size is larger,

Nit: Add "credential" before "store".

Make the same change to line 37 below.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:33: bool HasCredential(const HostPortPair&
origin);

Nit: document this method, even though it is self-explanatory.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:39: size_t AddCredential(const HostPortPair&
origin);

Is the return value useful to the caller?

It is strange that AddCredential() doesn't have a 'credential'
argument.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:42: size_t FindPosition(const HostPortPair&
origin);

Document FindPosition.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:44: static size_t npos;

This member name should end with an underscore '_'.

I believe npos can be const.

Nit: 'npos' is considered too short by our Style Guide.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:48: size_t last_added_;

Please document these data members.

The 'npos' member is especially hard to guess.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state_...
File net/spdy/spdy_credential_state_unittest.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state_...
net/spdy/spdy_credential_state_unittest.cc:14: }  // anonymous namespace

Delete lines 10-14.

[Ryan Hamilton, 2012-01-30 20:00:20]

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.cc
File net/socket/ssl_client_socket.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.cc...
net/socket/ssl_client_socket.cc:15: was_origin_bound_cert_sent_(false) {
On 2012/01/28 00:41:20, wtc wrote:
> 
> origin_bound_cert_type_ is not initialized.

Done.  (I have such a hard time remembering to do this because it happens by
default in Java and by default in C++ for object members).  Thanks.

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.h
File net/socket/ssl_client_socket.h (right):

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket.h#...
net/socket/ssl_client_socket.h:173: SSLClientCertType origin_bound_cert_type_;
On 2012/01/28 00:41:20, wtc wrote:
> 
> It seems that origin_bound_cert_type_ can replace the
> was_origin_bound_cert_sent_ field.
> 
> We can infer was_origin_bound_cert_sent_ by testing
> origin_bound_cert_type_ != CLIENT_CERT_INVALID_TYPE.

Good point.  Done.

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket_ns...
File net/socket/ssl_client_socket_nss.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/socket/ssl_client_socket_ns...
net/socket/ssl_client_socket_nss.cc:2725: }
On 2012/01/28 00:41:20, wtc wrote:
> 
> SSLClientSocketNSS needs to set origin_bound_cert_type_
> (which requires origin_bound_cert_type_ to be a protected
> member) or call set_origin_bound_cert_type().

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state.cc
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:17: last_added_(-1) {}
On 2012/01/28 00:41:20, wtc wrote:
> 
> Since npos and last_added_ are size_t, which is unsigned,
> it may be problematic to set them to -1.

I think using -1 is fine, since this is the approach that std::string uses:

http://www.cplusplus.com/reference/string/string/npos/

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:29: }
On 2012/01/28 00:41:20, wtc wrote:
> 
> In src/net, we usually omit curly braces in one-liner if
> statements.  Please follow this style in your code.
> 
> NOTE: Darin also made me use curly braces if there is an
> "else", even when both the "if" branch and "else" branch
> are one line.  For example,
>   if (condition) {
>     ++foo;
>   } else {
>     foo = 0;
>   }
> 
> This could be Darin's personal preference, so I recommend
> it but won't insist on it.

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:46: if (slots_[i].Equals(origin)) {
On 2012/01/28 00:41:20, wtc wrote:
> 
> Nit: It should be possible to replace this for loop with some
> STL technique.

Would that be better or simply different?  

I'm familiar with something like creating an iterator via begin(), incrementing
it and checking against end().  Is there also a mechanism for getting the index
from the iterator?

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:50: return -1;
On 2012/01/28 00:41:20, wtc wrote:
> 
> Change -1 to npos.

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:58: }
On 2012/01/28 00:41:20, wtc wrote:
> 
> I believe slots_.resize(size) can also be used to truncate.
> Could you find out if that's the case?

Done.  You're right, resize works just fine in this case.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state.h
File net/spdy/spdy_credential_state.h (right):

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:18: // make sure that we store the OBC
properly.
On 2012/01/28 00:41:20, wtc wrote:
> 
> This comment sounds like a TODO comment or a reminder for
> yourself.

Done. (Removed)

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:21: class NET_EXPORT_PRIVATE
SpdyCredentialState {
On 2012/01/28 00:41:20, wtc wrote:
> 
> Nit: it seems that this class should be named SpdyCredentialStore
> because you refer to "the store" in the comments below.

I changed the comment to eliminate the use of the word store.  I prefer not to
call this class a "Store" because it does not actually store the credentials. 
Rather, it tracks which origins have credentials associated with them. (Or
possibly "scopes" depending on where OBCs end up)

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:23: static size_t default_num_slots;
On 2012/01/28 00:41:20, wtc wrote:
> 
> The (static) data member should be listed after the methods.
> 
> The data member name should end with an underscore '_'.
> 
> This data member should be 'const'.  So it may need to
> follow the naming convention of constants.  Please check
> the Style Guide.

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:27: // Changes the size of the store.  If the
new size is larger,
On 2012/01/28 00:41:20, wtc wrote:
> 
> Nit: Add "credential" before "store".
> 
> Make the same change to line 37 below.

Reworked comments, though I removed reference to "store".

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:33: bool HasCredential(const HostPortPair&
origin);
On 2012/01/28 00:41:20, wtc wrote:
> 
> Nit: document this method, even though it is self-explanatory.

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:39: size_t AddCredential(const HostPortPair&
origin);
On 2012/01/28 00:41:20, wtc wrote:
> 
> Is the return value useful to the caller?

Yes, the caller uses that value to construct a CREDENTIAL frame.

> It is strange that AddCredential() doesn't have a 'credential'
> argument.

Perhaps I should rename the method to something like "SetHasCredential"?  Does
that work for you?  It doesn't take credentials because this class only records
that an origin *has* credentials.  It does not need to keep track of what they
are.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:42: size_t FindPosition(const HostPortPair&
origin);
On 2012/01/28 00:41:20, wtc wrote:
> 
> Document FindPosition.

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:44: static size_t npos;
On 2012/01/28 00:41:20, wtc wrote:
> 
> This member name should end with an underscore '_'.
> 
> I believe npos can be const.
> 
> Nit: 'npos' is considered too short by our Style Guide.

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.h:48: size_t last_added_;
On 2012/01/28 00:41:20, wtc wrote:
> 
> Please document these data members.
> 
> The 'npos' member is especially hard to guess.

Done.

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state_...
File net/spdy/spdy_credential_state_unittest.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state_...
net/spdy/spdy_credential_state_unittest.cc:14: }  // anonymous namespace
On 2012/01/28 00:41:20, wtc wrote:
> 
> Delete lines 10-14.

Done.

[wtc, 2012-01-31 19:21:25]

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state.cc
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/1/net/spdy/spdy_credential_state....
net/spdy/spdy_credential_state.cc:46: if (slots_[i].Equals(origin)) {

The STL technique I had in mind was the find() method.
But std::vector doesn't have the find() method.
It seems that we can use the find() *algorithm*
on std::vector, but I don't know how to get the index
from the returned iterator.  So this code is fine.

[wtc, 2012-01-31 22:13:19]

Patch Set 2 LGTM.

There are a lot of minor style issues.  Most of my comments
below are about these style nits and comment problems.

I didn't check some of the unit tests.  Please review them
yourself.

It would be nice to also have ramant review this CL.

http://codereview.chromium.org/9294001/diff/10001/net/base/net_log_event_type...
File net/base/net_log_event_type_list.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/base/net_log_event_type...
net/base/net_log_event_type_list.h:983: // Sending of a SPDY CREDENTIAL frame
(which send a certificate to the server).

Nit: send => sends

Note that in general the CREDENTIAL frame can send a
certificate chain, so the comments should reflect that.

http://codereview.chromium.org/9294001/diff/10001/net/socket/socket_test_util.cc
File net/socket/socket_test_util.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/socket/socket_test_util...
net/socket/socket_test_util.cc:690: }

Replace this while loop with a memset() call.

Return 'OK' instead of '0' (looks nicer).

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socket.h
File net/socket/ssl_client_socket.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket.h:145: // Origin Bound Certs are not supported.

"Socket" and "Origin Bound Certs" don't need to be capitalized.

Add "if" after "or NULL".

The original comments seem to avoid the "cert" abbreviation,
so I'd try to maintain that style in the new comments.

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket.h:152: virtual bool WasOriginBoundCertSent() const;

I think this method can remain lowercase Unix hacker style:
  was_origin_bound_cert_sent

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:47: }

Nit: remove curly braces from the if statements on
lines 45-47 and lines 55-57.

Please fix all such one-liner "if" and loop bodies
throughout this CL.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:56: last_added_ = num_slots_ - 1;

In keeping with the algorithm you used for eviction
on lines 33-37, it seems that last_added_ should be set to
0 here.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:28: // Returns true if there are credentials
associated with |origin|.

Nit: there are credentials => there is a credential ?

Or is "credential" usually used in plural form?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:42: // exists.

Nit: "if no entry exists" is confusing.  It can be misinterpreted
to mean slots_ is empty.

I suggest "if no entry for |origin| exists" or "if |origin| is not found".

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:51: size_t num_slots_;

How does num_slots_ differ from slots_.size()?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state_unittest.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state_unittest.cc:10:
//-----------------------------------------------------------------------------

Nit: this comment line is not necessary.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state_unittest.cc:44: TEST_F(SpdyCredentialStateTest,
HasCredentialsReturnsTrueWhenAdded) {

Nit: HasCredentials => HasCredential ?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state_unittest.cc:57: TEST_F(SpdyCredentialStateTest,
AddsReturnsPositionIfAlreadyInList) {

Nit: Adds => Add ?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_http_stream_u...
File net/spdy/spdy_http_stream_unittest.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_http_stream_u...
net/spdy/spdy_http_stream_unittest.cc:251: }

Nit: this can be done with a memset() call.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc
File net/spdy/spdy_session.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:20: #include "crypto/ec_private_key.h"

Do you know why we don't need to include "crypto/rsa_private_key.h"

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:80: dict->SetInteger("slost", slot_);

Typo: slost => slot

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:615: // Convert the string into a vector<unit8>

Nit: delete this comment?  It doesn't make sense to me.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:643: DCHECK(false);  // Error!

Use NOTREACHED().

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:653: credential.proof = proof_str;

Can you just do
  credential.proof.assign(proof.begin(), proof.end());
?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:655: CHECK(is_secure_);

Delete this line. Redundant with the DCHECK on line 601.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h
File net/spdy/spdy_session.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h#new...
net/spdy/spdy_session.h:121: int WriteCredentialFrame(std::string origin,

Document this method.

|origin| should be a const reference.  Also on lines 256
and 665.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h#new...
net/spdy/spdy_session.h:158: // Origin Bound Certs are not supported in this
session.

Nit: this class is named "SpdySession", so it's not clear
what "this Socket" refers to.

Add "if" after "or NULL".

Don't need to capitalize "Origin Bound Certs".

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h#new...
net/spdy/spdy_session.h:256: bool NeedsCredentials(HostPortPair origin);

Can this method be declared 'const'?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.cc
File net/spdy/spdy_stream.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.cc#new...
net/spdy/spdy_stream.cc:503: // Only OnWriteCompletee calls DoLoop(().

Can you fix this typo?  Completee => Complete

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.cc#new...
net/spdy/spdy_stream.cc:553: void SpdyStream::OnIOComplete(int result) {

The Style Guide recommends that this method be defined
before DoLoop, to match their declaration order in the
header file.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.h
File net/spdy/spdy_stream.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.h#newc...
net/spdy/spdy_stream.h:263: void OnIOComplete(int result);

Document this method.  It probably should be renamed to
suggest its relation to DoGetOriginBoundCertComplete().

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_test_util.h
File net/spdy/spdy_test_util.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_test_util.h#n...
net/spdy/spdy_test_util.h:163: // |credential| is the credential to sen.

Typo: sen => send

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_test_util.h#n...
net/spdy/spdy_test_util.h:165: spdy::SpdyFrame*
ConstructSpdyCredential(spdy::SpdyCredential credential);

|credential| should be a const reference.

Make the same change to the |settings| parameter on line 160
above.

[Ryan Hamilton, 2012-02-06 19:12:25]

http://codereview.chromium.org/9294001/diff/10001/net/base/net_log_event_type...
File net/base/net_log_event_type_list.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/base/net_log_event_type...
net/base/net_log_event_type_list.h:983: // Sending of a SPDY CREDENTIAL frame
(which send a certificate to the server).
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: send => sends
> 
> Note that in general the CREDENTIAL frame can send a
> certificate chain, so the comments should reflect that.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/socket/socket_test_util.cc
File net/socket/socket_test_util.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/socket/socket_test_util...
net/socket/socket_test_util.cc:690: }
On 2012/01/31 22:13:19, wtc wrote:
> 
> Replace this while loop with a memset() call.
> 
> Return 'OK' instead of '0' (looks nicer).

Done.  Duh!  Much cleaner.  Thanks.

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socket.h
File net/socket/ssl_client_socket.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket.h:145: // Origin Bound Certs are not supported.
On 2012/01/31 22:13:19, wtc wrote:
> 
> "Socket" and "Origin Bound Certs" don't need to be capitalized.
> 
> Add "if" after "or NULL".
> 
> The original comments seem to avoid the "cert" abbreviation,
> so I'd try to maintain that style in the new comments.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket.h:152: virtual bool WasOriginBoundCertSent() const;
On 2012/01/31 22:13:19, wtc wrote:
> 
> I think this method can remain lowercase Unix hacker style:
>   was_origin_bound_cert_sent

My understanding of the style guide is that we only use hacker case for methods
which are simple getters/settings for a member field of the same name.  That is
no longer the case for this method.  (That being said, I hate this convention
:>)

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:47: }
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: remove curly braces from the if statements on
> lines 45-47 and lines 55-57.
> 
> Please fix all such one-liner "if" and loop bodies
> throughout this CL.

Done.  Sorry, these braces were totally required in the google3 code I wrote and
I'm (obviously) having a hard time remembering that they are prohibited here.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:56: last_added_ = num_slots_ - 1;
On 2012/01/31 22:13:19, wtc wrote:
> 
> In keeping with the algorithm you used for eviction
> on lines 33-37, it seems that last_added_ should be set to
> 0 here.

I dont' think I understand why.  When we add an entry, we add it at the index
just after the last added.  In the case were the last added entry was the last
element in the array, we wrap to the front.

In the case where we make the array smaller, so that the index of the last added
element is no longer valid, we want to set last_added_ to the index of the last
element added, which would be the final element.  When we then call
SetHasCredential the next time, the new element will end up in slot 0, which I
think might be what you mean?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:28: // Returns true if there are credentials
associated with |origin|.
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: there are credentials => there is a credential ?
> 
> Or is "credential" usually used in plural form?

Done.  Yes, I think credentials is much more common than credential.  That being
said, since the name of the frame is CREDENTIAL, I think you're right to use the
singular form.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:42: // exists.
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: "if no entry exists" is confusing.  It can be misinterpreted
> to mean slots_ is empty.
> 
> I suggest "if no entry for |origin| exists" or "if |origin| is not found".

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:51: size_t num_slots_;
On 2012/01/31 22:13:19, wtc wrote:
> 
> How does num_slots_ differ from slots_.size()?

Looks like I misunderstandood the API here. vector::size() returns the actual
number of elements in the vector.  I understood this to mean that when the
vector was constructed with an initial capacity, but before any elements had
been explicitly added the size would be zero.  However, it looks like the vector
is initially populated with object constructed with the default constructor. 
I've removed this member variable.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state_unittest.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state_unittest.cc:10:
//-----------------------------------------------------------------------------
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: this comment line is not necessary.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state_unittest.cc:44: TEST_F(SpdyCredentialStateTest,
HasCredentialsReturnsTrueWhenAdded) {
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: HasCredentials => HasCredential ?

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state_unittest.cc:57: TEST_F(SpdyCredentialStateTest,
AddsReturnsPositionIfAlreadyInList) {
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: Adds => Add ?

Changed to "Set" since the method being tested was renamed from Add... to Set...

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_http_stream_u...
File net/spdy/spdy_http_stream_unittest.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_http_stream_u...
net/spdy/spdy_http_stream_unittest.cc:251: }
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: this can be done with a memset() call.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc
File net/spdy/spdy_session.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:20: #include "crypto/ec_private_key.h"
On 2012/01/31 22:13:19, wtc wrote:
> 
> Do you know why we don't need to include "crypto/rsa_private_key.h"

It is included in signature_creator.h, but I would be happy to include it here
if you wish?

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:80: dict->SetInteger("slost", slot_);
On 2012/01/31 22:13:19, wtc wrote:
> 
> Typo: slost => slot

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:615: // Convert the string into a vector<unit8>
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: delete this comment?  It doesn't make sense to me.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:643: DCHECK(false);  // Error!
On 2012/01/31 22:13:19, wtc wrote:
> 
> Use NOTREACHED().

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:653: credential.proof = proof_str;
On 2012/01/31 22:13:19, wtc wrote:
> 
> Can you just do
>   credential.proof.assign(proof.begin(), proof.end());
> ?

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:655: CHECK(is_secure_);
On 2012/01/31 22:13:19, wtc wrote:
> 
> Delete this line. Redundant with the DCHECK on line 601.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h
File net/spdy/spdy_session.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h#new...
net/spdy/spdy_session.h:121: int WriteCredentialFrame(std::string origin,
On 2012/01/31 22:13:19, wtc wrote:
> 
> Document this method.
> 
> |origin| should be a const reference.  Also on lines 256
> and 665.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h#new...
net/spdy/spdy_session.h:158: // Origin Bound Certs are not supported in this
session.
On 2012/01/31 22:13:19, wtc wrote:
> 
> Nit: this class is named "SpdySession", so it's not clear
> what "this Socket" refers to.
> 
> Add "if" after "or NULL".
> 
> Don't need to capitalize "Origin Bound Certs".

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.h#new...
net/spdy/spdy_session.h:256: bool NeedsCredentials(HostPortPair origin);
On 2012/01/31 22:13:19, wtc wrote:
> 
> Can this method be declared 'const'?

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.cc
File net/spdy/spdy_stream.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.cc#new...
net/spdy/spdy_stream.cc:503: // Only OnWriteCompletee calls DoLoop(().
On 2012/01/31 22:13:19, wtc wrote:
> 
> Can you fix this typo?  Completee => Complete
>   

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.cc#new...
net/spdy/spdy_stream.cc:553: void SpdyStream::OnIOComplete(int result) {
On 2012/01/31 22:13:19, wtc wrote:
> 
> The Style Guide recommends that this method be defined
> before DoLoop, to match their declaration order in the
> header file.

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.h
File net/spdy/spdy_stream.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_stream.h#newc...
net/spdy/spdy_stream.h:263: void OnIOComplete(int result);
On 2012/01/31 22:13:19, wtc wrote:
> 
> Document this method.  It probably should be renamed to
> suggest its relation to DoGetOriginBoundCertComplete().

I can't use that name because it it already taken.  I changed it to be
OnGetOriginBoundCertComplete.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_test_util.h
File net/spdy/spdy_test_util.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_test_util.h#n...
net/spdy/spdy_test_util.h:163: // |credential| is the credential to sen.
On 2012/01/31 22:13:19, wtc wrote:
> 
> Typo: sen => send

Done.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_test_util.h#n...
net/spdy/spdy_test_util.h:165: spdy::SpdyFrame*
ConstructSpdyCredential(spdy::SpdyCredential credential);
On 2012/01/31 22:13:19, wtc wrote:
> 
> |credential| should be a const reference.
> 
> Make the same change to the |settings| parameter on line 160
> above.

Done.

[wtc, 2012-02-06 22:48:55]

rch: let me respond to your comments first, and then I'll
review the new patch set.

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socket.h
File net/socket/ssl_client_socket.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket.h:152: virtual bool WasOriginBoundCertSent() const;

On 2012/02/06 19:12:25, Ryan Hamilton wrote:
>
> My understanding of the style guide is that we only use hacker case for
methods
> which are simple getters/settings for a member field of the same name.

In Chrome, this Style Guide recommendation also applies to
member functions that are as cheap as simple getters/setters.
That is the case for this method because its definition is
simply:

  return origin_bound_cert_type_ != CLIENT_CERT_INVALID_TYPE;

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:56: last_added_ = num_slots_ - 1;

On 2012/02/06 19:12:25, Ryan Hamilton wrote:
> 
> In the case where we make the array smaller, so that the index of the last
added
> element is no longer valid, we want to set last_added_ to the index of the
last
> element added, which would be the final element.  When we then call
> SetHasCredential the next time, the new element will end up in slot 0, which I
> think might be what you mean?

Ah, you are right.  What I wanted is that the new element
will end up in slot 0.  Your change accomplishes that.
Sorry about my confusion.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.h (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:51: size_t num_slots_;

On 2012/02/06 19:12:25, Ryan Hamilton wrote:
> However, it looks like the vector
> is initially populated with object constructed with the default constructor.

Yes, as far as I can tell, that's what the vector constructor
does.

To reserve an initial capacity without adding elements,
we need to use the reserve() method.

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc
File net/spdy/spdy_session.cc (right):

http://codereview.chromium.org/9294001/diff/10001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:20: #include "crypto/ec_private_key.h"

On 2012/02/06 19:12:25, Ryan Hamilton wrote:
>
> It is included in signature_creator.h, but I would be happy to include it here
> if you wish?

I suggest we do this:

1. Change signature_creator.h to use a forward declaration
of RSAPrivateKey instead of including "crypto/rsa_private_key.h".

2. Include "crypto/rsa_private_key.h" here.

I will take care of this change, because it may require
changing some other files.

[wtc, 2012-02-07 01:58:47]

Patch Set 7 LGTM.

In net/spdy/spdy_session.{h,cc} you added 'const' to some
input parameters, but you should also add '&' so that they
are const references.

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
File net/socket/ssl_client_socket_openssl.h (right):

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket_openssl.h:27: class OriginBoundCertService;

This forward declaration shouldn't be necessary because
"net/socket/ssl_client_socket.h" already has a forward
declaration of class OriginBoundCertService.

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket_openssl.h:59: virtual OriginBoundCertService*
GetOriginBoundCertService() const;

This should be moved to the "SSLClientSocket implementation"
section below.

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
File net/socket/ssl_client_socket_win.h (right):

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket_win.h:31: class OriginBoundCertService;

This forward declaration should be unnecessary.

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:15: : slots_(num_slots),

IMPORTANT: I think we should delete the
    slots_(num_slots),
initializer, and add a
  slots_.reserve(num_slots);
call to the constructor's body.

But this requires changing slots_.size() to
slots_.capacity() in some of the code below.

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc
File net/spdy/spdy_session.cc (right):

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:71: const std::string origin)

Can this be a const *reference* ?

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:555: bool SpdySession::NeedsCredentials(const
HostPortPair origin) const {

Can this be a const *reference* ?

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:606: int SpdySession::WriteCredentialFrame(const
std::string origin,

Can this be a const *reference* ?

[Ryan Hamilton, 2012-02-07 17:28:38]

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
File net/socket/ssl_client_socket_openssl.h (right):

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket_openssl.h:27: class OriginBoundCertService;
On 2012/02/07 01:58:47, wtc wrote:
> 
> This forward declaration shouldn't be necessary because
> "net/socket/ssl_client_socket.h" already has a forward
> declaration of class OriginBoundCertService.

Done.

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket_openssl.h:59: virtual OriginBoundCertService*
GetOriginBoundCertService() const;
On 2012/02/07 01:58:47, wtc wrote:
> 
> This should be moved to the "SSLClientSocket implementation"
> section below.

Done.

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
File net/socket/ssl_client_socket_win.h (right):

http://codereview.chromium.org/9294001/diff/25001/net/socket/ssl_client_socke...
net/socket/ssl_client_socket_win.h:31: class OriginBoundCertService;
On 2012/02/07 01:58:47, wtc wrote:
> 
> This forward declaration should be unnecessary.

Done.

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:15: : slots_(num_slots),
On 2012/02/07 01:58:47, wtc wrote:
> 
> IMPORTANT: I think we should delete the
>     slots_(num_slots),
> initializer, and add a
>   slots_.reserve(num_slots);
> call to the constructor's body.
> 
> But this requires changing slots_.size() to
> slots_.capacity() in some of the code below.

I think this would be a mistake.  Here is the documentation for reserve()

http://www.cplusplus.com/reference/stl/vector/reserve/

In particular:

"This informs the vector of a planned increase in size, although notice that the
parameter n informs of a minimum, so the resulting capacity may be any capacity
equal or larger than this."

The way I read this, if we switch to using capacity, it's possible that Chrome
will think there are a different number of slots available in the vector than
the server thinks.  (I considered this approach initially, but the documentation
persuaded me that it was not a good idea.)

What do you think?

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc
File net/spdy/spdy_session.cc (right):

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:71: const std::string origin)
On 2012/02/07 01:58:47, wtc wrote:
> 
> Can this be a const *reference* ?

Done.

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:555: bool SpdySession::NeedsCredentials(const
HostPortPair origin) const {
On 2012/02/07 01:58:47, wtc wrote:
> 
> Can this be a const *reference* ?

Done.

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_session.cc#ne...
net/spdy/spdy_session.cc:606: int SpdySession::WriteCredentialFrame(const
std::string origin,
On 2012/02/07 01:58:47, wtc wrote:
> 
> Can this be a const *reference* ?

Done.

[wtc, 2012-02-07 22:10:37]

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/25001/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:15: : slots_(num_slots),

My proposed change requires we be able to specify the vector's
capacity exactly.  So the reserve() method is inappropriate
since it merely specifies the minimum capacity.

[wtc, 2012-02-07 22:21:15]

Patch Set 9 LGTM.

http://codereview.chromium.org/9294001/diff/33029/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.cc (right):

http://codereview.chromium.org/9294001/diff/33029/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.cc:40: for (size_t i = 0; i < slots_.size(); i++)
{

The only problem with the current code is that when the'
slots_ vector has empty slots, this for loop will still
look for 'origin' in the empty slots.  Fortunately the
Equals comparison will fail because the empty slots have
empty host name, so we will still get correct results.

http://codereview.chromium.org/9294001/diff/33029/net/spdy/spdy_credential_st...
File net/spdy/spdy_credential_state.h (right):

http://codereview.chromium.org/9294001/diff/33029/net/spdy/spdy_credential_st...
net/spdy/spdy_credential_state.h:50: // Index of the last origin added to
|slots|.

Nit: |slots| => |slots_|

[commit-bot: I haz the power, 2012-02-08 00:35:43]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/9294001/44003

[commit-bot: I haz the power, 2012-02-08 02:45:40]

Change committed as 120923