Allow the analysis server to connect to the client over stdio.

Allow the analysis server to connect to the client over stdio.

R=brianwilkerson@google.com

Committed: https://code.google.com/p/dart/source/detail?r=35095

[Paul Berry, 2014-04-15 22:03:27]

Jaime: this change should start to unblock you from implementing the connection
from the editor to the analysis server.  A few caveats, though:

- There's currently no way to disable the HTTP interface.  Both HTTP and stdio
interfaces are active as soon as you start the analysis server.

- At the moment you need to pass the server the command line parameter "--port
NNN" to tell it what port to serve HTTP over.  That will become optional when I
get around to making HTTP optional.

- When the server starts it says "Listening on port ${port}" to standard output.
 I need to change this to standard error so that the editor doesn't try to read
it and interpret it as JSON.

- I plan to add a JSON message to the protocol which the server generates
immediately upon a successful connection, before receiving any requests from the
client.

- I think it would make sense for the server to exit when standard input is
closed.  That is not implemented yet, so when you are done with the server
you'll need to kill it with a signal.

[Brian Wilkerson, 2014-04-15 22:13:16]

LGTM

https://codereview.chromium.org/236133013/diff/1/pkg/analysis_server/lib/stdi...
File pkg/analysis_server/lib/stdio_server.dart (right):

https://codereview.chromium.org/236133013/diff/1/pkg/analysis_server/lib/stdi...
pkg/analysis_server/lib/stdio_server.dart:35: );
nit: the formatting here is rather unfortunate. Not sure what to recommend,
though.

[Paul Berry, 2014-04-15 22:43:07]

https://codereview.chromium.org/236133013/diff/1/pkg/analysis_server/lib/stdi...
File pkg/analysis_server/lib/stdio_server.dart (right):

https://codereview.chromium.org/236133013/diff/1/pkg/analysis_server/lib/stdi...
pkg/analysis_server/lib/stdio_server.dart:35: );
On 2014/04/15 22:13:16, Brian Wilkerson wrote:
> nit: the formatting here is rather unfortunate. Not sure what to recommend,
> though.

Yeah, that was dartfmt's decision.  I'm not thrilled with it either.

[Paul Berry, 2014-04-15 22:53:08]

Committed patchset #1 manually as r35095 (presubmit successful).