Fixes the tests written for http://codereview.chromium.org/3432009/show...

Fixes the tests written for http://codereview.chromium.org/3432009/show
to use DeterministicSocketData which should avoid the flaky test
problem that 3432009 introduced.

--

Add a new class SpdyProxyClientSocket which implements ClientSocket
by sending a CONNECT request via a SPDY SYN_STREAM frame to a SPDY proxy,
and then reading/writing data to/from SPDY Data frames.

BUG=29625
TEST=none


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=61144

[Ryan Hamilton, 2010-09-29 21:04:32]

Hi Guys,

I think/hope this fixes the flaky test that I committed as part of 3432009.  I
rewrote the tests to use DeterministicSocketData and that seems predictable. 
Can either of you take a look?

[Ryan Hamilton, 2010-09-30 18:05:49]

On 2010/09/29 21:04:32, Ryan Hamilton wrote:
> Hi Guys,
> 
> I think/hope this fixes the flaky test that I committed as part of 3432009.  I
> rewrote the tests to use DeterministicSocketData and that seems predictable. 
> Can either of you take a look?

I should mention that I also reorganized the tests so that the order made more
sense.  For some reason, codereview.chromium.org seems confused about the "Delta
from patch set 1" for the unit tests :(

[willchan no longer on Chromium, 2010-09-30 18:10:17]

I have say this issue made me not review it quickly.  Because I have to manually
do deltas myself, I punted this until I get a larger chunk of time available. 
If you can upload patchsets so it makes it easier to identify the delta, I can
do a quick review.  Otherwise, I'll take a look when I get a chance after perf.

On 2010/09/30 18:05:49, Ryan Hamilton wrote:
> On 2010/09/29 21:04:32, Ryan Hamilton wrote:
> > Hi Guys,
> > 
> > I think/hope this fixes the flaky test that I committed as part of 3432009. 
I
> > rewrote the tests to use DeterministicSocketData and that seems predictable.

> > Can either of you take a look?
> 
> I should mention that I also reorganized the tests so that the order made more
> sense.  For some reason, http://codereview.chromium.org seems confused about
the "Delta
> from patch set 1" for the unit tests :(

[Ryan Hamilton, 2010-09-30 19:35:12]

On 2010/09/30 18:10:17, willchan wrote:
> I have say this issue made me not review it quickly.  Because I have to
manually
> do deltas myself, I punted this until I get a larger chunk of time available. 
> If you can upload patchsets so it makes it easier to identify the delta, I can
> do a quick review.  Otherwise, I'll take a look when I get a chance after perf

Sorry about that!  I believe I have "fixed" the state of this patch.  You should
be able to look at the most recent patch set (#4) and diff that against the
previous (#3) to see the *actual* changes that this CL introduces on top of CL
3432009.

[willchan no longer on Chromium, 2010-09-30 23:28:18]

I don't quite understand the flaky test problem that you were seeing, but I'm
fine with all the changes I see here anyway.  LGTM.

On 2010/09/30 19:35:12, Ryan Hamilton wrote:
> On 2010/09/30 18:10:17, willchan wrote:
> > I have say this issue made me not review it quickly.  Because I have to
> manually
> > do deltas myself, I punted this until I get a larger chunk of time
available. 
> > If you can upload patchsets so it makes it easier to identify the delta, I
can
> > do a quick review.  Otherwise, I'll take a look when I get a chance after
perf
> 
> Sorry about that!  I believe I have "fixed" the state of this patch.  You
should
> be able to look at the most recent patch set (#4) and diff that against the
> previous (#3) to see the *actual* changes that this CL introduces on top of CL
> 3432009.

[Ryan Hamilton, 2010-10-01 00:34:00]

On 2010/09/30 23:28:18, willchan wrote:
> I don't quite understand the flaky test problem that you were seeing, but I'm
> fine with all the changes I see here anyway.  LGTM.

Great, thanks!  To be honest, I'm not sure I understand the flakiness either. 
However, I vandebo did have a reasonable complaint about the nature of tests
that depend on a read to block a write, or vice-versa.  In addition, I did have
a hard time getting OrderedSocketData to behave the way I wanted. 
OrderedSocketData has a 100ms wait which also makes me suspicious.  In any case,
I'll keep my fingers crossed :>