|
|
Boost Testing : |
From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2005-02-15 19:16:02
Robert Ramey writes:
> In updating the explicit markup for the serialization library I came up with
> a could of questions:
>
> a) what is the difference between "mark-expected-failures" and
> "mark-failure"?
<mark-failure> is a part of the "old" markup scheme where you had to
have a separate <test> entry for every marked test case, even if the
failures being marked shared the same note, e.g.
<test name="array3">
<mark-failure>
<toolset name="borland-5_6_4"/>
<toolset name="msvc"/>
<note refid="3"/>
</mark-failure>
<mark-failure>
<toolset name="sunpro"/>
<note refid="4"/>
</mark-failure>
</test>
<test name="array4">
<mark-failure>
<toolset name="borland-5_6_4"/>
<toolset name="msvc"/>
<note refid="3"/>
</mark-failure>
</test>
"mark-expected-failures" allows you to fold the above into
<mark-expected-failures>
<toolset name="borland-5_6_4"/>
<toolset name="msvc"/>
<test name="array4">
<test name="array3">
<note refid="3"/>
</mark-expected-failures>
<mark-expected-failures>
<toolset name="sunpro"/>
<test name="array3">
<note refid="4"/>
</mark-expected-failures>
> b) in reviewing the schema - I find that I can write <test...><mark-failure>
> but not <mark-failure><test...> .
Correct.
> On the other hand, I can write both
> <test...><mark-expected-failure> AND<mark-expected-failures><test...> .
Nope. Try to validate it.
> i)Is there a reason for this asymetry?
No, since there is no asymetry :).
> ii)Can it be corrected?
What would be the advantage of <mark-failure><test...> over the current
<mark-expected-failures> semantics?
-- Aleksey Gurtovoy MetaCommunications Engineering