Try to tolerate running on contexts with SIGPIPE enabled.

Try to tolerate running on contexts with SIGPIPE enabled.

Although Chromium wisely disables this ancient UNIX bug, the net stack is part
of Cronet which may be used in other consumers. Rather than modify the global
signal dispositions (bad manners for a library) or propagate this requirement
to the callers (also bad manners, if sound), make an attempt to suppress the
signal on a per-socket basis.

Unfortunately, iOS/OSX and Linux kindly have different APIs for this, so we
need to touch two places.

BUG=467132
Committed: https://crrev.com/281502752bdcee2f121b99d3a971301b8c271297
Cr-Commit-Position: refs/heads/master@{#398121}

[davidben, 2016-06-01 20:50:35]

davidben@chromium.org changed reviewers:
+ mmenke@chromium.org

[mmenke, 2016-06-01 21:45:11]

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_descripto...
File net/socket/socket_descriptor.cc (right):

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_descripto...
net/socket/socket_descriptor.cc:42: close(result);
closesocket

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_posix.cc
File net/socket/socket_posix.cc (right):

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_posix.cc#...
net/socket/socket_posix.cc:445: #if defined(OS_LINUX)
OS_LINUX excludes Android, I believe.  Should we also do this on Android?

[davidben, 2016-06-01 21:50:59]

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_descripto...
File net/socket/socket_descriptor.cc (right):

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_descripto...
net/socket/socket_descriptor.cc:42: close(result);
On 2016/06/01 21:45:11, mmenke wrote:
> closesocket

It's just close on UNIX. closesocket is a Windows-ism because Windows SOCKETs
are HANDLEs rather than file descriptors (which aren't HANDLEs and are merely a
C runtime fabrication).

But I seem to need a header.

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_posix.cc
File net/socket/socket_posix.cc (right):

https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_posix.cc#...
net/socket/socket_posix.cc:445: #if defined(OS_LINUX)
On 2016/06/01 21:45:11, mmenke wrote:
> OS_LINUX excludes Android, I believe.  Should we also do this on Android?

Ugh, so it does. Done.

[mmenke, 2016-06-01 22:12:15]

On 2016/06/01 21:50:59, davidben (OOO until 5-31) wrote:
>
https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_descripto...
> File net/socket/socket_descriptor.cc (right):
> 
>
https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_descripto...
> net/socket/socket_descriptor.cc:42: close(result);
> On 2016/06/01 21:45:11, mmenke wrote:
> > closesocket
> 
> It's just close on UNIX. closesocket is a Windows-ism because Windows SOCKETs
> are HANDLEs rather than file descriptors (which aren't HANDLEs and are merely
a
> C runtime fabrication).
> 
> But I seem to need a header.
> 
> https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_posix.cc
> File net/socket/socket_posix.cc (right):
> 
>
https://codereview.chromium.org/2034433002/diff/1/net/socket/socket_posix.cc#...
> net/socket/socket_posix.cc:445: #if defined(OS_LINUX)
> On 2016/06/01 21:45:11, mmenke wrote:
> > OS_LINUX excludes Android, I believe.  Should we also do this on Android?
> 
> Ugh, so it does. Done.

LGTM

[davidben, 2016-06-01 22:41:30]

The CQ bit was checked by davidben@chromium.org

[commit-bot: I haz the power, 2016-06-01 22:42:43]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2034433002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2034433002/20001

[commit-bot: I haz the power, 2016-06-02 01:55:46]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-02 01:55:47]

Try jobs failed on following builders:
  linux_chromium_rel_ng on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[davidben, 2016-06-06 19:24:21]

On 2016/06/02 01:55:47, commit-bot: I haz the power wrote:
> Try jobs failed on following builders:
>   linux_chromium_rel_ng on tryserver.chromium.linux (JOB_FAILED,
>
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

Oops, apparently this never landed.

[davidben, 2016-06-06 19:24:26]

The CQ bit was checked by davidben@chromium.org

[commit-bot: I haz the power, 2016-06-06 19:25:08]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2034433002/20001

[mmenke, 2016-06-06 19:25:34]

On 2016/06/06 19:24:21, davidben wrote:
> On 2016/06/02 01:55:47, commit-bot: I haz the power wrote:
> > Try jobs failed on following builders:
> >   linux_chromium_rel_ng on tryserver.chromium.linux (JOB_FAILED,
> >
>
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)
> 
> Oops, apparently this never landed.

GPU tests again.  We should just expect those to always fail, and if they pass,
consider it a bug.

[commit-bot: I haz the power, 2016-06-06 21:03:23]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-06-06 21:04:35]

Description was changed from

==========
Try to tolerate running on contexts with SIGPIPE enabled.

Although Chromium wisely disables this ancient UNIX bug, the net stack is part
of Cronet which may be used in other consumers. Rather than modify the global
signal dispositions (bad manners for a library) or propagate this requirement
to the callers (also bad manners, if sound), make an attempt to suppress the
signal on a per-socket basis.

Unfortunately, iOS/OSX and Linux kindly have different APIs for this, so we
need to touch two places.

BUG=467132
==========

to

==========
Try to tolerate running on contexts with SIGPIPE enabled.

Although Chromium wisely disables this ancient UNIX bug, the net stack is part
of Cronet which may be used in other consumers. Rather than modify the global
signal dispositions (bad manners for a library) or propagate this requirement
to the callers (also bad manners, if sound), make an attempt to suppress the
signal on a per-socket basis.

Unfortunately, iOS/OSX and Linux kindly have different APIs for this, so we
need to touch two places.

BUG=467132

Committed: https://crrev.com/281502752bdcee2f121b99d3a971301b8c271297
Cr-Commit-Position: refs/heads/master@{#398121}
==========

[commit-bot: I haz the power, 2016-06-06 21:04:36]

Patchset 2 (id:??) landed as
https://crrev.com/281502752bdcee2f121b99d3a971301b8c271297
Cr-Commit-Position: refs/heads/master@{#398121}