On 25 April 2013 03:31, Levente Uzonyi <leves at elte.hu> wrote:
> On Wed, 24 Apr 2013, Frank Shearar wrote:
>
>> So I wrote a very basic CommandLineToolSet (please! save me and
>> suggest a better name!). It can "debug" errors by dumping a stack
>> trace to stderr and kill the image.
>>
>> So now the CI tests use this, and the following tests fail:
>>
>> ToolsTests.Browser.BrowseTest.testBrowseClass
>> ToolsTests.Browser.BrowseTest.testBrowseHierarchyClass
>> ToolsTests.Browser.BrowseTest.testBrowseHierarchyInstance
>> ToolsTests.Browser.BrowseTest.testBrowseHierarchyMataclass
>> ToolsTests.Browser.BrowseTest.testBrowseInstance
>> ToolsTests.Browser.BrowseTest.testBrowseMetaclass
>> ToolsTests.Browser.DependencyBrowserTest.testBrowse
>> ToolsTests.Inspector.WeakSetInspectorTest.testSymbolTableM6812
>>
>> That's because these tests actually want the StandardToolSet. One idea
>> that occurs to me is to use a #setUp/#tearDown combo to re-set the
>> ToolSet. That comes at the risk of things blowing up in a test causing
>> CI jobs to hang; the very reason I wrote the CommandLineToolSet.
>>
>> So before I do that, any other ideas? Essentially I'm looking for a
>> scoped ToolSet change. I think?
>
>
> What about making CommandLineToolSet a subclass of StandardToolSet? That way
> everything would work unless there's an error.

Yes, I suppose I could do that. It would certainly scratch my current
itch. Thanks!

The other option is to fill out the ToolSet protocol with empty
methods. But that wouldn't be very least-surprise-ish.

frank

> Levente
>
>>
>> frank
>>
>>
>