Re: Blame ExpandEnvironmentStringsA for the failing NT4 test in shlwapi:shreg

[Alexandre Julliard]

Richard Cohen <richard(a)daijobu.co.uk> writes:
> This is failing because ExpandEnvironmentStringsA is returning twice
> the number of characters it should.
>
> I would like to mark this test (and some others) as todo(nt40), but
> the testing framework isn't flexible enough to let me do that (the
> only real options seem to be "wine" and "windows").
>
> Would a patch to extend the framework to allow a bit mask, something
> like todo(win9x|winxp), be acceptable?

A todo is supposed to mark something that will need to be addressed
later on. Obviously we can't fix the Windows code, so that's not a
todo, it's just a behavior of the API that should either be handled by
the test, or simply not tested at all if the results are not reliable.

-- 
Alexandre Julliard
julliard(a)winehq.org

Blame ExpandEnvironmentStringsA for the failing NT4 test in shlwapi:shreg

[Richard Cohen]

Alexandre Julliard wrote:
> A todo is supposed to mark something that will need to be addressed
> later on. Obviously we can't fix the Windows code, so that's not a
> todo, it's just a behavior of the API that should either be handled by
> the test, or simply not tested at all if the results are not reliable.

Yes, the API is not reliable, but we should make sure that Wine is doing 
  the *right thing*. The above choices make the test suite less useful 
for regression testing.

Surely, it is also good to document any bugs we find in Windows, and, 
though perhaps not for NT4, Microsoft *may* fix them in the future.

Why not rename "todo" to "bug_in" ?