From: Douglas Gregor (gregod_at_[hidden])
Date: 2002-02-10 19:24:25
On Sunday 10 February 2002 07:08 pm, you wrote:
> ----- Original Message -----
> From: "Douglas Gregor" <gregod_at_[hidden]>
> > Perhaps. If we were to require a green dashboard for all platforms before
> > release, and use "expected failures" when a compiler/platform cannot deal
> > with a certain testcase. Then a "diff" between the old and new expected
> > failures lists would tell what fails now that passed previously (and
> > won't likely be fixed).
> I think the biggest hurdle is that we need some way for library authors to
> specify which platforms they intend to support, and for suppliers of
> support for other platforms (when they are not the author) to do likewise.
I completely agree. Any ideas? The best I've thought of would be to
standardize the build names that describe the compiler/platform, e.g.,
and stick the build name into $(PLATFORM).
Then we add an "expected-failure" rule to the regression testing system, so
that the author can do something like:
case freebsd-3,*: expected-failure wregress ;
case winnt-msvc-*: expected-failure lambda_test ;
Doesn't quite scale to multiple build names in a single Jam invocation,
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk