From: Vladimir Prus (ghost_at_[hidden])
Date: 2002-05-03 10:57:35
David Abrahams wrote:
> > > Also, note that in many cases an un-normalized path is
> > > the
> > > same as its normalized version, so no tests will discover if you
> > > forget a conversion somewhere.
> > True. However, in case of unnormalized paths, tests that deal with
> > directory
> > names will be much harder to write. Or am I wrong?
> I think that as long as the path module provides the right rules, you're
> wrong. What specific cases did you have in mind?
For example, project-test1.py compares the output which contains project
location with stored results. If project location is output in an
os-dependent manner, the test won't. Or... can we use normalized
representation for debug output only?
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