Swallow errors in subscriptions.

Swallow errors in subscriptions.

BUG=https://code.google.com/p/dart/issues/detail?id=19815

[Bob Nystrom, 2014-07-09 00:28:05]

Anders, is this what you had in mind?

If so, I'll land this and then one of us can put together a Windows build to
send to our customer and see if it helps.

[nweiz, 2014-07-09 02:02:02]

https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
File pkg/http_multi_server/lib/src/utils.dart (right):

https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
pkg/http_multi_server/lib/src/utils.dart:22: // If one of the subscriptions has
an error (usually IPv6 failing late),
This comment should mention the tracking issue. It should also mention that this
implementation of mergeStreams is different from the implementation(s) elsewhere
in the repo.

https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
pkg/http_multi_server/lib/src/utils.dart:24: subscriptions.remove(subscription);
If this isn't the last subscription, cancel it as well.

https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/test/h...
File pkg/http_multi_server/test/http_multi_server_test.dart (right):

https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/test/h...
pkg/http_multi_server/test/http_multi_server_test.dart:136: }
I'm kind of uncomfortable about using different IPv6-detection logic in the
tests than we do in the library. If we can't trust this logic to be accurate
everywhere, how can we trust that the test won't be flaky?

[Bob Nystrom, 2014-07-10 20:05:27]

On 2014/07/09 02:02:02, nweiz wrote:
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
> File pkg/http_multi_server/lib/src/utils.dart (right):
> 
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
> pkg/http_multi_server/lib/src/utils.dart:22: // If one of the subscriptions
has
> an error (usually IPv6 failing late),
> This comment should mention the tracking issue. It should also mention that
this
> implementation of mergeStreams is different from the implementation(s)
elsewhere
> in the repo.
> 
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
> pkg/http_multi_server/lib/src/utils.dart:24:
subscriptions.remove(subscription);
> If this isn't the last subscription, cancel it as well.
> 
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/test/h...
> File pkg/http_multi_server/test/http_multi_server_test.dart (right):
> 
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/test/h...
> pkg/http_multi_server/test/http_multi_server_test.dart:136: }
> I'm kind of uncomfortable about using different IPv6-detection logic in the
> tests than we do in the library. If we can't trust this logic to be accurate
> everywhere, how can we trust that the test won't be flaky?

Marking this closed since you're taking it over, right?

- bob

[nweiz, 2014-07-10 20:58:15]

On 2014/07/10 20:05:27, Bob Nystrom wrote:
> On 2014/07/09 02:02:02, nweiz wrote:
> >
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
> > File pkg/http_multi_server/lib/src/utils.dart (right):
> > 
> >
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
> > pkg/http_multi_server/lib/src/utils.dart:22: // If one of the subscriptions
> has
> > an error (usually IPv6 failing late),
> > This comment should mention the tracking issue. It should also mention that
> this
> > implementation of mergeStreams is different from the implementation(s)
> elsewhere
> > in the repo.
> > 
> >
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/lib/sr...
> > pkg/http_multi_server/lib/src/utils.dart:24:
> subscriptions.remove(subscription);
> > If this isn't the last subscription, cancel it as well.
> > 
> >
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/test/h...
> > File pkg/http_multi_server/test/http_multi_server_test.dart (right):
> > 
> >
>
https://codereview.chromium.org/375113002/diff/1/pkg/http_multi_server/test/h...
> > pkg/http_multi_server/test/http_multi_server_test.dart:136: }
> > I'm kind of uncomfortable about using different IPv6-detection logic in the
> > tests than we do in the library. If we can't trust this logic to be accurate
> > everywhere, how can we trust that the test won't be flaky?
> 
> Marking this closed since you're taking it over, right?
> 
> - bob

Sure.