Add field trial stats for alternate_protocol.

Add field trial stats for alternate_protocol. The histogram we collected are: 
http stats when alternate protocol is available but npn was not negotiated (
which means not usng spdy), and when alternate is available AND spdy is used.

Noticable changes:
1. In http_network_transaction.cc, changed the logic that always parse 
response from server for alternate protocol and remember that in 
HttpAlternateProtocols strucuture. We need to remember this to collect 
stats for servers with alternate protocol support but used http for.
2. In spdy_stream.cc, get rid of the response copy from spdy_stream. This
copy overwrites some early status in response set in http_network_transaction.


TEST=http_network_transaction_unittest.cc
BUG=46689
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=50927

[lzheng, 2010-06-16 18:17:19]

This won't build yet, because we need to update a patch to webkit. But the logic
should all be there and my build and test works locally.

Mike, can you do me a favor to patch the webkit again? I am working on my local
webkit repository and it is still not working.

[lzheng, 2010-06-18 20:39:13]

I thought I did but obviously I missed the reviewers when I sent out this for
review. 

===========
This won't build yet, because we need to update a patch to webkit. Mike kindly
offers to help me patch the webkit with the change I sent to him.

[mbelshe, 2010-06-19 00:07:41]

Hi Lei -

I'll get your patch to webkit done over the weekend.

Sorry for being a bit slow

Mike


On Fri, Jun 18, 2010 at 1:39 PM, <lzheng@chromium.org> wrote:

> I thought I did but obviously I missed the reviewers when I sent out this
> for
> review.
>
> ===========
> This won't build yet, because we need to update a patch to webkit. Mike
> kindly
> offers to help me patch the webkit with the change I sent to him.
>
>
>
>
> http://codereview.chromium.org/2808010/show
>

[willchan no longer on Chromium, 2010-06-19 00:29:15]

http://codereview.chromium.org/2808010/diff/42001/43005
File chrome/renderer/render_view.cc (right):

http://codereview.chromium.org/2808010/diff/42001/43005#newcode4691
chrome/renderer/render_view.cc:4691: // Spdy requests triggered by alternate
protocol is excluded here.
s/is/are/

http://codereview.chromium.org/2808010/diff/42001/43005#newcode4692
chrome/renderer/render_view.cc:4692: // This is because when alternate protocol
is avaiable, field trial will
s/field trial/FieldTrial/g

http://codereview.chromium.org/2808010/diff/42001/43005#newcode4693
chrome/renderer/render_view.cc:4693: // either use npn_spdy or pure http, not
npn_http via SLT. That will cause
SLT?  Do you mean TLS?

http://codereview.chromium.org/2808010/diff/42001/43010
File net/spdy/spdy_session.cc (right):

http://codereview.chromium.org/2808010/diff/42001/43010#newcode967
net/spdy/spdy_session.cc:967: HttpResponseInfo* response =
stream->GetMutableResponseInfo();
Hm, this isn't quite going to work.  ukai's moving out all the http specific
stuff out of SpdySession.  Is there any way to do this in the transaction code
instead?

[lzheng1, 2010-06-19 07:09:41]

On Fri, Jun 18, 2010 at 5:29 PM, <willchan@chromium.org> wrote:

>
> http://codereview.chromium.org/2808010/diff/42001/43005
> File chrome/renderer/render_view.cc (right):
>
> http://codereview.chromium.org/2808010/diff/42001/43005#newcode4691
> chrome/renderer/render_view.cc:4691: // Spdy requests triggered by
> alternate protocol is excluded here.
> s/is/are/
>
> http://codereview.chromium.org/2808010/diff/42001/43005#newcode4692
> chrome/renderer/render_view.cc:4692: // This is because when alternate
> protocol is avaiable, field trial will
> s/field trial/FieldTrial/g
>
> http://codereview.chromium.org/2808010/diff/42001/43005#newcode4693
> chrome/renderer/render_view.cc:4693: // either use npn_spdy or pure
> http, not npn_http via SLT. That will cause
> SLT?  Do you mean TLS?
>
> http://codereview.chromium.org/2808010/diff/42001/43010
> File net/spdy/spdy_session.cc (right):
>
> http://codereview.chromium.org/2808010/diff/42001/43010#newcode967
> net/spdy/spdy_session.cc:967: HttpResponseInfo* response =
> stream->GetMutableResponseInfo();
> Hm, this isn't quite going to work.  ukai's moving out all the http
> specific stuff out of SpdySession.  Is there any way to do this in the
> transaction code instead?
>
> Ukai actually did what I wanted. He passed the pointer of response from
http_network_transaction to spdy_http_stream instead of creating a totally
new one here and later copying back (which will overwrites any information
before the response is passed to spdy_http_stream). I could wait after he
submits and my change will be simpler (at least no need to change
spdy_session.cc).



>
> http://codereview.chromium.org/2808010/show
>

[lzheng1, 2010-06-22 21:25:47]

+ Ukai

Will:

Your comments are addressed. Another look?

Ukai:
Looks like your change http://codereview.chromium.org/2667002/show was
reverted. When will you plan to submit? Do you mind if I get this CL in
first if yours will take a couple of more days? Your merge with my cl should
be simple: you just need to ignore the changes I made in spdy_stream.xx and
spdy_session.xx. Since I am trying to pass the pointer of response_ from
http_transaction down to stream and session instead of create a new response
in SpdySession::Respond() as we are doing now. You did what I wanted.

Thanks!

Lei
On Sat, Jun 19, 2010 at 12:09 AM, Lei Zheng <lzheng@google.com> wrote:

>
>
> On Fri, Jun 18, 2010 at 5:29 PM, <willchan@chromium.org> wrote:
>
>>
>> http://codereview.chromium.org/2808010/diff/42001/43005
>> File chrome/renderer/render_view.cc (right):
>>
>> http://codereview.chromium.org/2808010/diff/42001/43005#newcode4691
>> chrome/renderer/render_view.cc:4691: // Spdy requests triggered by
>> alternate protocol is excluded here.
>> s/is/are/
>>
>> http://codereview.chromium.org/2808010/diff/42001/43005#newcode4692
>> chrome/renderer/render_view.cc:4692: // This is because when alternate
>> protocol is avaiable, field trial will
>> s/field trial/FieldTrial/g
>>
>> http://codereview.chromium.org/2808010/diff/42001/43005#newcode4693
>> chrome/renderer/render_view.cc:4693: // either use npn_spdy or pure
>> http, not npn_http via SLT. That will cause
>> SLT?  Do you mean TLS?
>>
>> http://codereview.chromium.org/2808010/diff/42001/43010
>> File net/spdy/spdy_session.cc (right):
>>
>> http://codereview.chromium.org/2808010/diff/42001/43010#newcode967
>> net/spdy/spdy_session.cc:967: HttpResponseInfo* response =
>> stream->GetMutableResponseInfo();
>> Hm, this isn't quite going to work.  ukai's moving out all the http
>> specific stuff out of SpdySession.  Is there any way to do this in the
>> transaction code instead?
>>
>> Ukai actually did what I wanted. He passed the pointer of response from
> http_network_transaction to spdy_http_stream instead of creating a totally
> new one here and later copying back (which will overwrites any information
> before the response is passed to spdy_http_stream). I could wait after he
> submits and my change will be simpler (at least no need to change
> spdy_session.cc).
>
>
>
>>
>> http://codereview.chromium.org/2808010/show
>>
>
>

[lzheng, 2010-06-25 06:55:45]

I integrated the change with ErikChen's check in. The change is ready for review
now. 

The webkit CL is still not landed yet. However, if we could get the CL ready,
once webkit is done, I could check it in to make it available for next dev
channel push.

[willchan no longer on Chromium, 2010-06-25 17:20:21]

LGTM

[Mike Belshe, 2010-06-25 17:57:22]

lgtm