Add IssueAuthToken support to the TokenService.

The token service should now support issuing auth tokens.
The service will fetch tokens in the background and broadcast notifications after it's done.
Move some common constants out into a new file.
Modify TestNotificationTracker to support subclassing as it doesn't properly make a deep copy of notifications on the stack.
TEST=unit tests included
BUG=47093
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=54028

[tim (not reviewing), 2010-07-27 19:12:17]

http://codereview.chromium.org/3024002/diff/17001/18004
File chrome/browser/net/gaia/token_service.h (right):

http://codereview.chromium.org/3024002/diff/17001/18004#newcode41
chrome/browser/net/gaia/token_service.h:41: TokenRequestFailedDetails() {}
why's this needed?

http://codereview.chromium.org/3024002/diff/17001/18004#newcode46
chrome/browser/net/gaia/token_service.h:46: const GaiaAuthError& error() const {
return error_; }
Any chance you can switch this to use GoogleServiceAuthError now?  I asked
before, but haven't seen the follow up change yet :)  If we could at least end
the divergence, that would be great.

http://codereview.chromium.org/3024002/diff/17001/18004#newcode65
chrome/browser/net/gaia/token_service.h:65: // On login, StartFetchingTokens()
should be kicked off.
'be called to kick off the process' or something?

http://codereview.chromium.org/3024002/diff/17001/18004#newcode66
chrome/browser/net/gaia/token_service.h:66: // Tokens will be fetched for all
services in the background.
which services?

http://codereview.chromium.org/3024002/diff/17001/18004#newcode69
chrome/browser/net/gaia/token_service.h:69: bool HasTokenForService(const char*
const service);
const ?

http://codereview.chromium.org/3024002/diff/17001/18004#newcode80
chrome/browser/net/gaia/token_service.h:80: bool AreCredentialsValid();
const ?

http://codereview.chromium.org/3024002/diff/17001/18004#newcode90
chrome/browser/net/gaia/token_service.h:90: };
DISALLOW_COPY_AND_ASSIGN(TokenService)

http://codereview.chromium.org/3024002/diff/17001/18013
File chrome/common/notification_type.h (right):

http://codereview.chromium.org/3024002/diff/17001/18013#newcode1028
chrome/common/notification_type.h:1028: // these notifications is issued per new
token.
explain what details / source are

http://codereview.chromium.org/3024002/diff/17001/18013#newcode1031
chrome/common/notification_type.h:1031: // If a token request failed, one of
these is issued per failed request.
explain what details / source are

[chron_chromium.org, 2010-07-27 20:08:28]

http://codereview.chromium.org/3024002/diff/17001/18004
File chrome/browser/net/gaia/token_service.h (right):

http://codereview.chromium.org/3024002/diff/17001/18004#newcode41
chrome/browser/net/gaia/token_service.h:41: TokenRequestFailedDetails() {}
On 2010/07/27 19:12:17, timsteele wrote:
> why's this needed?
If you mean the constructor, tests need to use it to make a request to assign
things to.
If you mean the details struct, it's for CAPTCHA.

http://codereview.chromium.org/3024002/diff/17001/18004#newcode46
chrome/browser/net/gaia/token_service.h:46: const GaiaAuthError& error() const {
return error_; }
On 2010/07/27 19:12:17, timsteele wrote:
> Any chance you can switch this to use GoogleServiceAuthError now?  I asked
> before, but haven't seen the follow up change yet :)  If we could at least end
> the divergence, that would be great.

It's a little bit involved. ChromeOS code has some parsing logic in it that
depends on reading the response entirely. I really should remove the parsing
logic from ChromeOS code and have it rely entirely on the error struct, but
doing so will take some time. For the time being I had de prioritized it in
favor of getting more functionality in. When we start wiring in the NTP captcha
stuff we'll have to do this anyway.

If you'd prefer, I can prioritize that and do it next before the webdb
serialization.

http://codereview.chromium.org/3024002/diff/17001/18004#newcode65
chrome/browser/net/gaia/token_service.h:65: // On login, StartFetchingTokens()
should be kicked off.
On 2010/07/27 19:12:17, timsteele wrote:
> 'be called to kick off the process' or something? 

Done.

http://codereview.chromium.org/3024002/diff/17001/18004#newcode66
chrome/browser/net/gaia/token_service.h:66: // Tokens will be fetched for all
services in the background.
On 2010/07/27 19:12:17, timsteele wrote:
> which services?

Done.

http://codereview.chromium.org/3024002/diff/17001/18004#newcode69
chrome/browser/net/gaia/token_service.h:69: bool HasTokenForService(const char*
const service);
On 2010/07/27 19:12:17, timsteele wrote:
> const ?

Done.

http://codereview.chromium.org/3024002/diff/17001/18004#newcode80
chrome/browser/net/gaia/token_service.h:80: bool AreCredentialsValid();
On 2010/07/27 19:12:17, timsteele wrote:
> const ?

Done.

http://codereview.chromium.org/3024002/diff/17001/18004#newcode90
chrome/browser/net/gaia/token_service.h:90: };
On 2010/07/27 19:12:17, timsteele wrote:
> DISALLOW_COPY_AND_ASSIGN(TokenService)

Done.

http://codereview.chromium.org/3024002/diff/17001/18013
File chrome/common/notification_type.h (right):

http://codereview.chromium.org/3024002/diff/17001/18013#newcode1028
chrome/common/notification_type.h:1028: // these notifications is issued per new
token.
On 2010/07/27 19:12:17, timsteele wrote:
> explain what details / source are

Done.

http://codereview.chromium.org/3024002/diff/17001/18013#newcode1031
chrome/common/notification_type.h:1031: // If a token request failed, one of
these is issued per failed request.
On 2010/07/27 19:12:17, timsteele wrote:
> explain what details / source are

Done.

[chron, 2010-07-27 20:37:54]

I've filed a bug http://code.google.com/p/chromium/issues/detail?id=50408 to
track the unification of the error structs. I'll get my ChromeOS setup
warmed up while I work on another issue and gut it sooner.

[tim (not reviewing), 2010-07-27 20:44:36]

in the interest of productivity, LGTM, though please read my comment.  thanks
for filing the bug; I just think the sooner you tackle it the easier it will be.

http://codereview.chromium.org/3024002/diff/17001/18004
File chrome/browser/net/gaia/token_service.h (right):

http://codereview.chromium.org/3024002/diff/17001/18004#newcode46
chrome/browser/net/gaia/token_service.h:46: const GaiaAuthError& error() const {
return error_; }
On 2010/07/27 20:08:28, chron_chromium.org wrote:
> On 2010/07/27 19:12:17, timsteele wrote:
> > Any chance you can switch this to use GoogleServiceAuthError now?  I asked
> > before, but haven't seen the follow up change yet :)  If we could at least
end
> > the divergence, that would be great.
> 
> It's a little bit involved. ChromeOS code has some parsing logic in it that
> depends on reading the response entirely. I really should remove the parsing
> logic from ChromeOS code and have it rely entirely on the error struct, but
> doing so will take some time. For the time being I had de prioritized it in
> favor of getting more functionality in. When we start wiring in the NTP
captcha
> stuff we'll have to do this anyway.
> 
> If you'd prefer, I can prioritize that and do it next before the webdb
> serialization.

I just thought it might help to stop plumbing the struct through, e.g. by making
the notification send the GSAE instead.  I can't see how whatever code is
listening to this notification can have vastly different parsing code since at
the end of the day the two containers really just give you two or three pieces
of POD.