Re: [Boost-testing] Testing with libraries that fail to build

From: Robert Ramey (ramey_at_[hidden])
Date: 2005-06-24 01:17:07

Jonathan Turkanis wrote:
> Robert Ramey wrote:
>> Libraries that are not header only have special problems regarding
>> the tests
>>
>> a) if he library fails to build - all the tests are marked "fail"
>> This is very misleading as a test cannot fail if it is not run.
>
> Marking a test as failed when it fails to build doesn't seem
> misleading to me, although it would be convenient if I could tell at
> a glance whether there was a compiler error, linker error or runtime
> failure.

Its marking a test as failed when the LIBRARY  fails to build thats
misleading. So one issue to resolved shows as 250 test failures.  To me,
that's misleading

>
>> b) a library may fail to build because of another library fails to
>> build. A common case is where the cannot run because the Test library
>> fails to build for the indicated toolset.  Again all the tests are
>> marked "fail".  An this for failure to build code not even in the
>> library being tested !!
>
> I agree this is extremely frustrating. Ideally, failures should be
> counted as unresolved issues for a library only if all the
> prerequisite libraries built successfully.
>
> Jonathan