Rip out the last dart:io dependency from pkg/http.

Rip out the last dart:io dependency from pkg/http.

It now loads dart:io via mirrors if it's available. This allows it to be
compatible with the browser, so a new library "package:http/browser_client.dart"
is added to support that.

The use of mirrors was tested to ensure that it doesn't add any additional
filesize to dart2js compilations over what would be used if "dart:io" support
didn't exist.

R=rnystrom@google.com

Committed: https://code.google.com/p/dart/source/detail?r=35756

[nweiz, 2014-05-01 00:44:05]

Per Kevin's request, I'll hold off on submitting this until he gives the
go-ahead.

[Bob Nystrom, 2014-05-01 18:30:12]

https://codereview.chromium.org/261763002/diff/1/pkg/collection/lib/equality....
File pkg/collection/lib/equality.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/collection/lib/equality....
pkg/collection/lib/equality.dart:18: factory Equality() => const
DefaultEquality();
What's up with this?

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
File pkg/http/lib/browser_client.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
pkg/http/lib/browser_client.dart:79: new ClientException("XMLHttpRequest
error.", request.url),
Probably want to include the original error here.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
pkg/http/lib/browser_client.dart:91: void close() {
Stream.close() and other close() methods return Futures. Should this?

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/src/io.dart
File pkg/http/lib/src/io.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/src/io.dart#new...
pkg/http/lib/src/io.dart:1: // Copyright (c) 2014, the Dart project authors. 
Please see the AUTHORS file
Doc comments on the public stuff in here.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/src/io.dart#new...
pkg/http/lib/src/io.dart:13: bool get supported => _library != null;
Nit, but how about putting all of the private fields together and having this
above or below them all?

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status
File pkg/pkg.status (right):

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status#newcode207
pkg/pkg.status:207: http/test/io/*: Fail, OK # Uses dart:io.
Skip?

[nweiz, 2014-05-01 19:34:54]

https://codereview.chromium.org/261763002/diff/1/pkg/collection/lib/equality....
File pkg/collection/lib/equality.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/collection/lib/equality....
pkg/collection/lib/equality.dart:18: factory Equality() => const
DefaultEquality();
On 2014/05/01 18:30:12, Bob Nystrom wrote:
> What's up with this?

Oops, this works around a dart2js bug but shouldn't be part of this CL. Removed.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
File pkg/http/lib/browser_client.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
pkg/http/lib/browser_client.dart:79: new ClientException("XMLHttpRequest
error.", request.url),
On 2014/05/01 18:30:12, Bob Nystrom wrote:
> Probably want to include the original error here.

There is no original error. The parameter name is misleading; it's actually just
a ProgressEvent with no metadata about the error that occurred. The XHR object
just doesn't expose that information for some reason.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
pkg/http/lib/browser_client.dart:91: void close() {
On 2014/05/01 18:30:12, Bob Nystrom wrote:
> Stream.close() and other close() methods return Futures. Should this?

The Client interface was created before that convention existed, but it should
probably be updated. That's an issue for a different CL, though.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/src/io.dart
File pkg/http/lib/src/io.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/src/io.dart#new...
pkg/http/lib/src/io.dart:1: // Copyright (c) 2014, the Dart project authors. 
Please see the AUTHORS file
On 2014/05/01 18:30:12, Bob Nystrom wrote:
> Doc comments on the public stuff in here.

Done.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/src/io.dart#new...
pkg/http/lib/src/io.dart:13: bool get supported => _library != null;
On 2014/05/01 18:30:12, Bob Nystrom wrote:
> Nit, but how about putting all of the private fields together and having this
> above or below them all?

Done.

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status
File pkg/pkg.status (right):

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status#newcode207
pkg/pkg.status:207: http/test/io/*: Fail, OK # Uses dart:io.
On 2014/05/01 18:30:12, Bob Nystrom wrote:
> Skip?

All the other tests use "Fail, OK" here. I don't understand the intricacies of
status files, but I figure being consistent is a safe bet.

[Bob Nystrom, 2014-05-01 19:52:14]

I'm curious what the right thing to do in the status file is, but otherwise
LGTM.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
File pkg/http/lib/browser_client.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
pkg/http/lib/browser_client.dart:79: new ClientException("XMLHttpRequest
error.", request.url),
On 2014/05/01 19:34:54, nweiz wrote:
> On 2014/05/01 18:30:12, Bob Nystrom wrote:
> > Probably want to include the original error here.
> 
> There is no original error. The parameter name is misleading; it's actually
just
> a ProgressEvent with no metadata about the error that occurred. The XHR object
> just doesn't expose that information for some reason.

Oh, heh. In that case, leave a comment explaining this.

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
pkg/http/lib/browser_client.dart:91: void close() {
On 2014/05/01 19:34:54, nweiz wrote:
> On 2014/05/01 18:30:12, Bob Nystrom wrote:
> > Stream.close() and other close() methods return Futures. Should this?
> 
> The Client interface was created before that convention existed, but it should
> probably be updated. That's an issue for a different CL, though.

SGTM.

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status
File pkg/pkg.status (right):

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status#newcode207
pkg/pkg.status:207: http/test/io/*: Fail, OK # Uses dart:io.
On 2014/05/01 19:34:54, nweiz wrote:
> On 2014/05/01 18:30:12, Bob Nystrom wrote:
> > Skip?
> 
> All the other tests use "Fail, OK" here. I don't understand the intricacies of
> status files, but I figure being consistent is a safe bet.

Well, it's inconsistent with how you handle http/test/html. Maybe ask on
dart-team@?

[nweiz, 2014-05-01 20:54:10]

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
File pkg/http/lib/browser_client.dart (right):

https://codereview.chromium.org/261763002/diff/1/pkg/http/lib/browser_client....
pkg/http/lib/browser_client.dart:79: new ClientException("XMLHttpRequest
error.", request.url),
On 2014/05/01 19:52:14, Bob Nystrom wrote:
> On 2014/05/01 19:34:54, nweiz wrote:
> > On 2014/05/01 18:30:12, Bob Nystrom wrote:
> > > Probably want to include the original error here.
> > 
> > There is no original error. The parameter name is misleading; it's actually
> just
> > a ProgressEvent with no metadata about the error that occurred. The XHR
object
> > just doesn't expose that information for some reason.
> 
> Oh, heh. In that case, leave a comment explaining this.

Done.

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status
File pkg/pkg.status (right):

https://codereview.chromium.org/261763002/diff/1/pkg/pkg.status#newcode207
pkg/pkg.status:207: http/test/io/*: Fail, OK # Uses dart:io.
On 2014/05/01 19:52:14, Bob Nystrom wrote:
> On 2014/05/01 19:34:54, nweiz wrote:
> > On 2014/05/01 18:30:12, Bob Nystrom wrote:
> > > Skip?
> > 
> > All the other tests use "Fail, OK" here. I don't understand the intricacies
of
> > status files, but I figure being consistent is a safe bet.
> 
> Well, it's inconsistent with how you handle http/test/html. Maybe ask on
> dart-team@?

Done.

[nweiz, 2014-05-05 18:30:43]

The crashing dart2js bug has been fixed. Kevin, am I good to go on submitting
this?

[Bob Nystrom, 2014-05-05 19:15:57]

Still LGTM.

[nweiz, 2014-05-05 19:36:41]

Committed patchset #3 manually as r35756 (presubmit successful).

[kasperl, 2014-05-06 08:53:13]

We should really, really, really avoid using mirrors for these things. It has
bitten us in the past and it's going to continue biting us in the future.