Staus file update: analyzer_experimental/test/ast_test seems to be passing

Staus file update: analyzer_experimental/test/ast_test seems to be passing

R=ricow@google.com

Committed: https://code.google.com/p/dart/source/detail?r=25917

[kustermann, 2013-08-08 07:57:15]

Seems to be passing since r25902.

[ricow1, 2013-08-08 08:14:53]

lgtm

[kustermann, 2013-08-08 08:19:50]

Committed patchset #1 manually as r25917 (presubmit successful).

[ahe, 2013-08-08 11:33:49]

DBC,  as I pointed out in https://codereview.chromium.org/22285004/, the test is
actually failing because it uses dart:io.

However, when we test with compiler=dart2js and runtime=d8, we add a special
option to dart2js that switches to the so-called mock-compilation mode. In this
mode, we generate all the code as we normally do, but remove the call to main.
This is useful for testing that dart2js can analyze dart:io programs (which
makes dart2dart more stable).

So because we don't call main, the generated program is a no-op.

Fortunately, we still have test coverage, because when we test on browsers, we
don't use "mock" compilation. However, the analyzer team has decided to mark all
analyzer tests as failing on browsers. This means that you cannot use the
analyzer in a web browser. I find that a bit surprising, and I wonder if this
decision was coordinated with Devon.

I also wonder if this decision was coordinated with Dan and Dan, as this affects
how the analyzer API can be used by programs like dartdoc.

Let me reiterate that when you don't run tests (or mark them as failing), you
don't have test coverage. When you don't have test coverage, your users will
start experiencing regressions and generally perceive your software as
unreliable and buggy.

So right now the situation is this: the package analyzer_experimental is *only*
for "server" applications.

[scheglov, 2013-08-08 16:44:04]

On 2013/08/08 11:33:49, ahe wrote:

OK, tests  for ast/scanner/parser will be restored with this CL.
https://codereview.chromium.org/22325013

> Let me reiterate that when you don't run tests (or mark them as failing), you
> don't have test coverage. When you don't have test coverage, your users will
> start experiencing regressions and generally perceive your software as
> unreliable and buggy.
> 
> So right now the situation is this: the package analyzer_experimental is
*only*
> for "server" applications.