Move signal_blocker to platform and use it by default in TEMP_FAILURE_RETRY.

Move signal_blocker to platform and use it by default in TEMP_FAILURE_RETRY.

BUG=http://code.google.com/p/dart/issues/detail?id=16927
R=johnmccutchan@google.com, sgjesse@google.com

Committed: https://code.google.com/p/dart/source/detail?r=33745

[Anders Johnsen, 2014-02-20 14:42:18]

This is only Linux for now - let me know what you think.

I took the time to go through man-pages to see what calls actually can return
EINTR - not that many it turns out :)

[Søren Gjesse, 2014-02-20 15:09:09]

https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
File runtime/bin/dbg_connection_linux.cc (right):

https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
runtime/bin/dbg_connection_linux.cc:73: int result = pipe(wakeup_fds_);
Should we add a NO_RETRY_EXPECTED macro to sanity-check this?

One that in DEBUG mode checks that EINTR is never returned, and does nothing in
release mode.

The man-pages could be inaccurate...?

[Anders Johnsen, 2014-02-20 15:10:16]

https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
File runtime/bin/dbg_connection_linux.cc (right):

https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
runtime/bin/dbg_connection_linux.cc:73: int result = pipe(wakeup_fds_);
On 2014/02/20 15:09:09, Søren Gjesse wrote:
> Should we add a NO_RETRY_EXPECTED macro to sanity-check this?
> 
> One that in DEBUG mode checks that EINTR is never returned, and does nothing
in
> release mode.
> 
> The man-pages could be inaccurate...?

Good idea - will into that tomorrow.

[Cutch, 2014-02-20 15:20:01]

https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
File runtime/bin/dbg_connection_linux.cc (right):

https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
runtime/bin/dbg_connection_linux.cc:73: int result = pipe(wakeup_fds_);
On 2014/02/20 15:09:09, Søren Gjesse wrote:
> Should we add a NO_RETRY_EXPECTED macro to sanity-check this?
> 
> One that in DEBUG mode checks that EINTR is never returned, and does nothing
in
> release mode.
> 
> The man-pages could be inaccurate...?

+1

[Cutch, 2014-03-11 19:42:10]

On 2014/02/20 15:20:01, Cutch wrote:
>
https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
> File runtime/bin/dbg_connection_linux.cc (right):
> 
>
https://codereview.chromium.org/165723007/diff/1/runtime/bin/dbg_connection_l...
> runtime/bin/dbg_connection_linux.cc:73: int result = pipe(wakeup_fds_);
> On 2014/02/20 15:09:09, Søren Gjesse wrote:
> > Should we add a NO_RETRY_EXPECTED macro to sanity-check this?
> > 
> > One that in DEBUG mode checks that EINTR is never returned, and does nothing
> in
> > release mode.
> > 
> > The man-pages could be inaccurate...?
> 
> +1

*ping*

[Anders Johnsen, 2014-03-14 09:07:43]

PTAL

[Cutch, 2014-03-14 14:40:38]

lgtm w/ comment

https://codereview.chromium.org/165723007/diff/110001/runtime/platform/signal...
File runtime/platform/signal_blocker.h (right):

https://codereview.chromium.org/165723007/diff/110001/runtime/platform/signal...
runtime/platform/signal_blocker.h:72: // This macro can be used to insert checks
that a call is not made, that
"call is not made" -> "call is made"

[Anders Johnsen, 2014-03-17 13:35:33]

PTAL, compiled and tested on Linux and Mac OS.

https://codereview.chromium.org/165723007/diff/110001/runtime/platform/signal...
File runtime/platform/signal_blocker.h (right):

https://codereview.chromium.org/165723007/diff/110001/runtime/platform/signal...
runtime/platform/signal_blocker.h:72: // This macro can be used to insert checks
that a call is not made, that
On 2014/03/14 14:40:39, Cutch wrote:
> "call is not made" -> "call is made"

Done.

[Søren Gjesse, 2014-03-17 13:41:44]

LGTM - there was quite a lot of NO_RETRY_EXPECTED.

[Anders Johnsen, 2014-03-17 13:51:45]

Committed patchset #6 manually as r33745 (presubmit successful).