From: David Abrahams (dave_at_[hidden])
Date: 2002-11-12 10:44:22
Vladimir Prus <ghost_at_[hidden]> writes:
> I actually have access to win2000 box (only it's slow, and
> free bcc32 is the only compiler). So I went ahead with
> project_test3 failed as well -- I was using "rm" on NT. I've
> klu^H^H^Hfixed that as well. There's a problem: variables in
> actions are looked up only in global scope, so I had to poke
> RM varaible in global scope, which is somewhat ugly.
We should fix that.
> project_test4 failed too, but the reason is different. I belive
> that your changes to error formatting broke it, because it tests
> that a certain error message is produced.
> I think we should discuss what do do with error messages. If we
> decide to use 'errors.error', then we should add an option to omit
> backtrace, so that we can check the error message only.
> I also think that if we don't come to a conclusion fast enought, the
> change must be (temporary) reverted, because developing with failing
> tests is hard.
I agree that the tests have to be resolved. However, I strongly
believe in having a single error entry point.
-- David Abrahams dave_at_[hidden] * http://www.boost-consulting.com Boost support, enhancements, training, and commercial distribution
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk