Re: [boost] status/Jamfile

From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-03-24 11:05:46

----- Original Message -----
From: "John Maddock" <john_maddock_at_[hidden]>

> Dave,
>
> I have some comments on the new status makefile/regression tests for
you.

Thanks!

> 1) I like the regression test code in status/Jamfile - is there any
chance
> of making this generally available as part of the boost build system -
it
> looks like the sort of thing that can be usefully reused.

That was part of the plan; Joerg Walter was working on it for a while,
but then I made a whole bunch of changes and I think he got discouraged
or ran out of time. However, please see this thread at jamboost:
http://groups.yahoo.com/group/jamboost/message/514

> 2) I don't think that the tests in status/Jamfile should be included
by
> default from the top level Jamfile - for one thing if a user just
wants to
> build the boost libraries, they may not want to stick around while
hundreds
> of test cases build (something that can take some time), for another
the
> current status/Jamfile implementation results in quite a number of
compiler
> errors scrolling past the screen (see below).

I agree. It can also cause a long delay before building starts just
because building the dependencies for the tests is expensive.

> 3) The current implementation of the link-fail and compile-fail tests
> results in a number of "failure" messages scrolling past the screen -
the
> link-fail test is the worst, but compile-fail isn't too good either.
The
> problem is that although these tests succeed, they none the less
indicate
> one or more failures occurred, for example the output from
> static_assert_test_8 is shown below (this is a successful test but you
> wouldn't necessarily know that from the output!):

Yep; fixing this would require some more core Jam extensions than I'm
ready to implement right now. The key, when you want to know what
failed, is to look for the "***********" lines in the output.

-Dave