Removed use of XmppSocketAdapter by sync.

Removed use of XmppSocketAdapter by sync.

Moved XmppSocketAdapter and friends to remoting directory.

Removed some dead code in jingle/.

BUG=54146
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=59012

[akalin, 2010-09-09 03:51:51]

+sergeyu for remoting stuff, +tim for everything else

[akalin, 2010-09-09 03:55:03]

Sergey, let me know if you guys would like to use ChromeAsyncSocket, too.  If
so, we can do that in a separate CL.

On 2010/09/09 03:51:51, akalin wrote:
> +sergeyu for remoting stuff, +tim for everything else

[akalin, 2010-09-09 18:16:31]

Fixed checkdeps problem.

On 2010/09/09 03:55:03, akalin wrote:
> Sergey, let me know if you guys would like to use ChromeAsyncSocket, too.  If
> so, we can do that in a separate CL.
> 
> On 2010/09/09 03:51:51, akalin wrote:
> > +sergeyu for remoting stuff, +tim for everything else

[Sergey Ulanov, 2010-09-09 18:23:09]

On 2010/09/09 03:55:03, akalin wrote:
> Sergey, let me know if you guys would like to use ChromeAsyncSocket,
> too.  If so, we can do that in a separate CL.

We do want to switch to Chromium sockets, but it is not possible yet because we
still use libjingle sockets for p2p connections, i.e. we still have to use
libjingle thread. We will be ready to switch after I implement socket server for
chromium sockets.
Remoting changes LGTM.

[akalin, 2010-09-09 22:10:05]

tim, ping!

On 2010/09/09 18:23:09, sergeyu wrote:
> On 2010/09/09 03:55:03, akalin wrote:
> > Sergey, let me know if you guys would like to use ChromeAsyncSocket,
> > too.  If so, we can do that in a separate CL.
> 
> We do want to switch to Chromium sockets, but it is not possible yet because
we
> still use libjingle sockets for p2p connections, i.e. we still have to use
> libjingle thread. We will be ready to switch after I implement socket server
for
> chromium sockets.
> Remoting changes LGTM.

[tim (not reviewing), 2010-09-09 22:44:35]

LGTM

http://codereview.chromium.org/3298021/diff/17001/18002
File chrome/browser/sync/tools/sync_listen_notifications.cc (right):

http://codereview.chromium.org/3298021/diff/17001/18002#newcode95
chrome/browser/sync/tools/sync_listen_notifications.cc:95: ssl_config, 4096, 64
* 1024, NULL);
would be nice if these numbers were constants

[akalin, 2010-09-09 22:47:34]

http://codereview.chromium.org/3298021/diff/17001/18002
File chrome/browser/sync/tools/sync_listen_notifications.cc (right):

http://codereview.chromium.org/3298021/diff/17001/18002#newcode95
chrome/browser/sync/tools/sync_listen_notifications.cc:95: ssl_config, 4096, 64
* 1024, NULL);
Will do, as separate CL.

On 2010/09/09 22:44:35, timsteele wrote:
> would be nice if these numbers were constants

[willchan no longer on Chromium, 2010-09-10 16:39:16]

Hi,

Apparently this changelist leads to writes to the ChromeNetLog (a thread-unsafe
object) from multiple threads, which is very bad (I'm hitting a ton of DCHECKs
in my debug linux Chromium build, which look to be happening on sync threads). 
I've reverted it locally on my chromium branch which fixes the DCHECKs.  I'm
going to revert it on trunk.  Sorry for any inconvenience.  I believe the
solution will be to make ChromeNetLog threadsafe.  I'm filing a bug for it too.

On 2010/09/09 22:47:34, akalin wrote:
> http://codereview.chromium.org/3298021/diff/17001/18002
> File chrome/browser/sync/tools/sync_listen_notifications.cc (right):
> 
> http://codereview.chromium.org/3298021/diff/17001/18002#newcode95
> chrome/browser/sync/tools/sync_listen_notifications.cc:95: ssl_config, 4096,
64
> * 1024, NULL);
> Will do, as separate CL.
> 
> On 2010/09/09 22:44:35, timsteele wrote:
> > would be nice if these numbers were constants

[akalin, 2010-09-10 17:08:38]

Hunh that is strange, given the fact that ChromeAsyncSocket is used with or
without this CL. All this CL does (ostensibly) is remove the now-unused
branches.
On Sep 10, 2010 9:39 AM, <willchan@chromium.org> wrote:
> Hi,
>
> Apparently this changelist leads to writes to the ChromeNetLog (a
> thread-unsafe
> object) from multiple threads, which is very bad (I'm hitting a ton of
> DCHECKs
> in my debug linux Chromium build, which look to be happening on sync
> threads).
> I've reverted it locally on my chromium branch which fixes the DCHECKs.
I'm
> going to revert it on trunk. Sorry for any inconvenience. I believe the
> solution will be to make ChromeNetLog threadsafe. I'm filing a bug for it
> too.
>
> On 2010/09/09 22:47:34, akalin wrote:
>> http://codereview.chromium.org/3298021/diff/17001/18002
>> File chrome/browser/sync/tools/sync_listen_notifications.cc (right):
>
>> http://codereview.chromium.org/3298021/diff/17001/18002#newcode95
>> chrome/browser/sync/tools/sync_listen_notifications.cc:95: ssl_config,
>> 4096,
> 64
>> * 1024, NULL);
>> Will do, as separate CL.
>
>> On 2010/09/09 22:44:35, timsteele wrote:
>> > would be nice if these numbers were constants
>
>
>
> http://codereview.chromium.org/3298021/show

[willchan no longer on Chromium, 2010-09-10 17:10:44]

Yeah, I read through the CL to try to understand why, but I don't know.
 There's a chance that I reverted the wrong changelist, but when I reverted,
this definitely was fixed (or maybe I just needed to restart).

On Fri, Sep 10, 2010 at 10:08 AM, Fred Akalin <akalin@chromium.org> wrote:

> Hunh that is strange, given the fact that ChromeAsyncSocket is used with or
> without this CL. All this CL does (ostensibly) is remove the now-unused
> branches.
> On Sep 10, 2010 9:39 AM, <willchan@chromium.org> wrote:
> > Hi,
> >
> > Apparently this changelist leads to writes to the ChromeNetLog (a
> > thread-unsafe
> > object) from multiple threads, which is very bad (I'm hitting a ton of
> > DCHECKs
> > in my debug linux Chromium build, which look to be happening on sync
> > threads).
> > I've reverted it locally on my chromium branch which fixes the DCHECKs.
> I'm
> > going to revert it on trunk. Sorry for any inconvenience. I believe the
> > solution will be to make ChromeNetLog threadsafe. I'm filing a bug for it
>
> > too.
> >
> > On 2010/09/09 22:47:34, akalin wrote:
> >> http://codereview.chromium.org/3298021/diff/17001/18002
> >> File chrome/browser/sync/tools/sync_listen_notifications.cc (right):
> >
> >> http://codereview.chromium.org/3298021/diff/17001/18002#newcode95
> >> chrome/browser/sync/tools/sync_listen_notifications.cc:95: ssl_config,
> >> 4096,
> > 64
> >> * 1024, NULL);
> >> Will do, as separate CL.
> >
> >> On 2010/09/09 22:44:35, timsteele wrote:
> >> > would be nice if these numbers were constants
> >
> >
> >
> > http://codereview.chromium.org/3298021/show
>

[willchan no longer on Chromium, 2010-09-10 17:13:50]

Btw, the other thing that I noticed happened was we were doing OCSP on these
other threads.  I wonder if the fake ssl adapter you're using is somehow
triggering OCSP checks?  That might be a bug.

On Fri, Sep 10, 2010 at 10:10 AM, William Chan (陈智昌)
<willchan@chromium.org>wrote:

> Yeah, I read through the CL to try to understand why, but I don't know.
>  There's a chance that I reverted the wrong changelist, but when I reverted,
> this definitely was fixed (or maybe I just needed to restart).
>
>
> On Fri, Sep 10, 2010 at 10:08 AM, Fred Akalin <akalin@chromium.org> wrote:
>
>> Hunh that is strange, given the fact that ChromeAsyncSocket is used with
>> or without this CL. All this CL does (ostensibly) is remove the now-unused
>> branches.
>> On Sep 10, 2010 9:39 AM, <willchan@chromium.org> wrote:
>> > Hi,
>> >
>> > Apparently this changelist leads to writes to the ChromeNetLog (a
>> > thread-unsafe
>> > object) from multiple threads, which is very bad (I'm hitting a ton of
>> > DCHECKs
>> > in my debug linux Chromium build, which look to be happening on sync
>> > threads).
>> > I've reverted it locally on my chromium branch which fixes the DCHECKs.
>> I'm
>> > going to revert it on trunk. Sorry for any inconvenience. I believe the
>> > solution will be to make ChromeNetLog threadsafe. I'm filing a bug for
>> it
>> > too.
>> >
>> > On 2010/09/09 22:47:34, akalin wrote:
>> >> http://codereview.chromium.org/3298021/diff/17001/18002
>> >> File chrome/browser/sync/tools/sync_listen_notifications.cc (right):
>> >
>> >> http://codereview.chromium.org/3298021/diff/17001/18002#newcode95
>> >> chrome/browser/sync/tools/sync_listen_notifications.cc:95: ssl_config,
>> >> 4096,
>> > 64
>> >> * 1024, NULL);
>> >> Will do, as separate CL.
>> >
>> >> On 2010/09/09 22:44:35, timsteele wrote:
>> >> > would be nice if these numbers were constants
>> >
>> >
>> >
>> > http://codereview.chromium.org/3298021/show
>>
>
>

[akalin, 2010-09-10 17:51:01]

The fake SSL adapter shouldn't be doing any real SSL work, so it
shouldn't be triggering OCSP checks.  However, I thought it was a
long-standing problem that the OCSP initialization wasn't thread-safe
etc., and we were using it on the sync xmpp thread (with or without
ChromeAsyncSocket).

On Fri, Sep 10, 2010 at 10:13 AM, William Chan (陈智昌)
<willchan@chromium.org> wrote:
> Btw, the other thing that I noticed happened was we were doing OCSP on these
> other threads.  I wonder if the fake ssl adapter you're using is somehow
> triggering OCSP checks?  That might be a bug.
>
> On Fri, Sep 10, 2010 at 10:10 AM, William Chan (陈智昌) <willchan@chromium.org>
> wrote:
>>
>> Yeah, I read through the CL to try to understand why, but I don't know.
>>  There's a chance that I reverted the wrong changelist, but when I reverted,
>> this definitely was fixed (or maybe I just needed to restart).
>>
>> On Fri, Sep 10, 2010 at 10:08 AM, Fred Akalin <akalin@chromium.org> wrote:
>>>
>>> Hunh that is strange, given the fact that ChromeAsyncSocket is used with
>>> or without this CL. All this CL does (ostensibly) is remove the now-unused
>>> branches.
>>>
>>> On Sep 10, 2010 9:39 AM, <willchan@chromium.org> wrote:
>>> > Hi,
>>> >
>>> > Apparently this changelist leads to writes to the ChromeNetLog (a
>>> > thread-unsafe
>>> > object) from multiple threads, which is very bad (I'm hitting a ton of
>>> > DCHECKs
>>> > in my debug linux Chromium build, which look to be happening on sync
>>> > threads).
>>> > I've reverted it locally on my chromium branch which fixes the DCHECKs.
>>> > I'm
>>> > going to revert it on trunk. Sorry for any inconvenience. I believe the
>>> > solution will be to make ChromeNetLog threadsafe. I'm filing a bug for
>>> > it
>>> > too.
>>> >
>>> > On 2010/09/09 22:47:34, akalin wrote:
>>> >> http://codereview.chromium.org/3298021/diff/17001/18002
>>> >> File chrome/browser/sync/tools/sync_listen_notifications.cc (right):
>>> >
>>> >> http://codereview.chromium.org/3298021/diff/17001/18002#newcode95
>>> >> chrome/browser/sync/tools/sync_listen_notifications.cc:95: ssl_config,
>>> >> 4096,
>>> > 64
>>> >> * 1024, NULL);
>>> >> Will do, as separate CL.
>>> >
>>> >> On 2010/09/09 22:44:35, timsteele wrote:
>>> >> > would be nice if these numbers were constants
>>> >
>>> >
>>> >
>>> > http://codereview.chromium.org/3298021/show
>>
>
>