Revert "Make stdout/stderr async"

Revert "Make stdout/stderr async"

This makes writes to the stdout and stderr IOSinks in dart:io blocking again.

To get the non-blocking behaviour the getter "nonBlocking" will return an
IOSink with real non-blocking behaviour.

Updated the class exposed for stderr to also support checking for whether it
is connected to a terminal and checking the terminal size.

R=ajohnsen@google.com, iposva@google.com

BUG=

Committed: https://code.google.com/p/dart/source/detail?r=42909

[Søren Gjesse, 2014-12-17 15:34:03]

sgjesse@google.com changed reviewers:
+ iposva@google.com

[Ivan Posva, 2014-12-18 08:09:32]

I strongly oppose re-introducing this irregularity into the IO system.

Also I do not understand why Console has to be removed even if the decision is
being made to introduce blocking IO for stdout/stderr by default.

-Ivan

[Søren Gjesse, 2014-12-19 13:55:22]

On 2014/12/18 08:09:32, Ivan Posva wrote:
> I strongly oppose re-introducing this irregularity into the IO system.
> 
> Also I do not understand why Console has to be removed even if the decision is
> being made to introduce blocking IO for stdout/stderr by default.
> 
> -Ivan

If we go back to blocking IO for stdout/stderr, then I see no point in having
console as well. That will just make two different ways to do blocking writes to
stdout/stderr. I think users will have enough difficulties in figuring out
whether to use stdout or stdout.nonBlocking.

The decision to "keep" stdout/stderr blocking has been made.

[Søren Gjesse, 2015-01-09 08:21:45]

sgjesse@google.com changed reviewers:
+ ajohnsen@google.com

[Søren Gjesse, 2015-01-09 08:21:46]

Anders can you please take a look?

[Anders Johnsen, 2015-01-09 12:35:11]

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart
File sdk/lib/io/stdio.dart (right):

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:157: * In some situations this blocking behavior is
undesirable as it does no
no -> not

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:200: if (_fd == null) return _sink;
when is _fd null?

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:223: _StdConsumer(int fd) : _file =
_File._openStdioSync(fd);
Perhaps '_open*' is the wrong name, as the fd is already open.

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:225: Future addStream(Stream<List<int>> stream) {
This makes me wonder if

void main() {
  stdout.writeln("foo");
  print("bar");
}

will put bar before foo. Can you confirm foo is first?

https://codereview.chromium.org/807833002/diff/20001/tests/standalone/io/cons...
File tests/standalone/io/console_test.dart (left):

https://codereview.chromium.org/807833002/diff/20001/tests/standalone/io/cons...
tests/standalone/io/console_test.dart:20:
Expect.equals('stdout\r\n\r\ntuodts\r\nABCDEFGHIJKLM\r\n', result.stdout);
use .nonBlocking and keep this test?

[Søren Gjesse, 2015-01-14 10:12:21]

PTAL

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart
File sdk/lib/io/stdio.dart (right):

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:157: * In some situations this blocking behavior is
undesirable as it does no
On 2015/01/09 12:35:11, Anders Johnsen wrote:
> no -> not

Done.

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:200: if (_fd == null) return _sink;
On 2015/01/09 12:35:11, Anders Johnsen wrote:
> when is _fd null?

Should never be. Removed.

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:223: _StdConsumer(int fd) : _file =
_File._openStdioSync(fd);
On 2015/01/09 12:35:11, Anders Johnsen wrote:
> Perhaps '_open*' is the wrong name, as the fd is already open.

Sure, but these are the names we already used. I can change open to get in a
separate CL.

https://codereview.chromium.org/807833002/diff/20001/sdk/lib/io/stdio.dart#ne...
sdk/lib/io/stdio.dart:225: Future addStream(Stream<List<int>> stream) {
On 2015/01/09 12:35:11, Anders Johnsen wrote:
> This makes me wonder if
> 
> void main() {
>   stdout.writeln("foo");
>   print("bar");
> }
> 
> will put bar before foo. Can you confirm foo is first?

It prints

foo
bar

This is basically the code which was there before I changed it to non-blocking.

https://codereview.chromium.org/807833002/diff/20001/tests/standalone/io/cons...
File tests/standalone/io/console_test.dart (left):

https://codereview.chromium.org/807833002/diff/20001/tests/standalone/io/cons...
tests/standalone/io/console_test.dart:20:
Expect.equals('stdout\r\n\r\ntuodts\r\nABCDEFGHIJKLM\r\n', result.stdout);
On 2015/01/09 12:35:11, Anders Johnsen wrote:
> use .nonBlocking and keep this test?

Done, added as stdio_nonblocking_test.dert.

[Anders Johnsen, 2015-01-14 13:56:59]

LGTM

https://codereview.chromium.org/807833002/diff/40001/sdk/lib/_internal/compil...
File sdk/lib/_internal/compiler/js_lib/io_patch.dart (right):

https://codereview.chromium.org/807833002/diff/40001/sdk/lib/_internal/compil...
sdk/lib/_internal/compiler/js_lib/io_patch.dart:188: static _lock(int id, int
lock, int start, int end) {
Unrelated?

[Søren Gjesse, 2015-01-15 14:15:20]

https://codereview.chromium.org/807833002/diff/40001/sdk/lib/_internal/compil...
File sdk/lib/_internal/compiler/js_lib/io_patch.dart (right):

https://codereview.chromium.org/807833002/diff/40001/sdk/lib/_internal/compil...
sdk/lib/_internal/compiler/js_lib/io_patch.dart:188: static _lock(int id, int
lock, int start, int end) {
On 2015/01/14 13:56:59, Anders Johnsen wrote:
> Unrelated?

Yes, added after rebase.

[Søren Gjesse, 2015-01-15 14:18:32]

Committed patchset #3 (id:40001) manually as 42909 (presubmit successful).