Add bufferOutput to HttpServer, HttpResponse and HttpClientRequest.

Add bufferOutput to HttpServer, HttpResponse and HttpClientRequest.

Setting bufferOutput to false will disable our internal buffering, and
write the buffers directly to the socket.

BUG=https://code.google.com/p/dart/issues/detail?id=14953
R=sgjesse@google.com

Committed: https://code.google.com/p/dart/source/detail?r=34821

[Søren Gjesse, 2014-04-08 11:54:25]

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http.dart
File sdk/lib/io/http.dart (right):

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http.dart#newcode189
sdk/lib/io/http.dart:189: bool bufferOutput;
Do we need this default? I assume this will only be set to false for specific
URLs.

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http_impl.dart
File sdk/lib/io/http_impl.dart (right):

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http_impl.dart#ne...
sdk/lib/io/http_impl.dart:414: bool bufferOutput;
What happens if this is changed while the body is being written?

Should we only support it being changed before the first byte of the body is
written? If not it should flush the current buffer when turned off.

Should compression be turned off if this is set to false?

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http_impl.dart#ne...
sdk/lib/io/http_impl.dart:1183: add(new Uint8List.view(_buffer.buffer, 0,
_length));
Please add a comment on why you are doing this.

https://codereview.chromium.org/228453002/diff/1/tests/standalone/io/http_cli...
File tests/standalone/io/http_client_connect_test.dart (right):

https://codereview.chromium.org/228453002/diff/1/tests/standalone/io/http_cli...
tests/standalone/io/http_client_connect_test.dart:167: /*
Remove these comments.

https://codereview.chromium.org/228453002/diff/1/tests/standalone/io/http_cli...
tests/standalone/io/http_client_connect_test.dart:175:
testGetDataServerForceClose();
There should be a test where the server writes several times and the client
receives separate chunks. Maybe have the client complete a future in the server
between chunks.

[Anders Johnsen, 2014-04-08 12:40:07]

PTAL

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http.dart
File sdk/lib/io/http.dart (right):

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http.dart#newcode189
sdk/lib/io/http.dart:189: bool bufferOutput;
On 2014/04/08 11:54:25, Søren Gjesse wrote:
> Do we need this default? I assume this will only be set to false for specific
> URLs.

Good point. Will remove this one.

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http_impl.dart
File sdk/lib/io/http_impl.dart (right):

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http_impl.dart#ne...
sdk/lib/io/http_impl.dart:414: bool bufferOutput;
On 2014/04/08 11:54:25, Søren Gjesse wrote:
> What happens if this is changed while the body is being written?
> 
> Should we only support it being changed before the first byte of the body is
> written? If not it should flush the current buffer when turned off.
> 
> Should compression be turned off if this is set to false?

Done. Sadly, this is needed, as zlib buffers as well.

https://codereview.chromium.org/228453002/diff/1/sdk/lib/io/http_impl.dart#ne...
sdk/lib/io/http_impl.dart:1183: add(new Uint8List.view(_buffer.buffer, 0,
_length));
On 2014/04/08 11:54:25, Søren Gjesse wrote:
> Please add a comment on why you are doing this.

Done.

https://codereview.chromium.org/228453002/diff/1/tests/standalone/io/http_cli...
File tests/standalone/io/http_client_connect_test.dart (right):

https://codereview.chromium.org/228453002/diff/1/tests/standalone/io/http_cli...
tests/standalone/io/http_client_connect_test.dart:167: /*
On 2014/04/08 11:54:25, Søren Gjesse wrote:
> Remove these comments.

Done.

https://codereview.chromium.org/228453002/diff/1/tests/standalone/io/http_cli...
tests/standalone/io/http_client_connect_test.dart:175:
testGetDataServerForceClose();
On 2014/04/08 11:54:25, Søren Gjesse wrote:
> There should be a test where the server writes several times and the client
> receives separate chunks. Maybe have the client complete a future in the
server
> between chunks.

Done.

[Søren Gjesse, 2014-04-08 13:03:23]

lgtm

https://codereview.chromium.org/228453002/diff/10001/tests/standalone/io/http...
File tests/standalone/io/http_client_connect_test.dart (right):

https://codereview.chromium.org/228453002/diff/10001/tests/standalone/io/http...
tests/standalone/io/http_client_connect_test.dart:197:
response.writeln('output$count');
He he the client writes to the response.

[Anders Johnsen, 2014-04-08 13:06:59]

https://codereview.chromium.org/228453002/diff/10001/tests/standalone/io/http...
File tests/standalone/io/http_client_connect_test.dart (right):

https://codereview.chromium.org/228453002/diff/10001/tests/standalone/io/http...
tests/standalone/io/http_client_connect_test.dart:197:
response.writeln('output$count');
On 2014/04/08 13:03:24, Søren Gjesse wrote:
> He he the client writes to the response.

The server response, yeah.

[Anders Johnsen, 2014-04-08 13:07:30]

Committed patchset #2 manually as r34821 (presubmit successful).