From: Vladimir Prus (ghost_at_[hidden])
Date: 2004-10-26 04:37:31
> how should "current-disk" relative native paths be handled, e.g.:
> "\foo\bar\giz.h" (windows), "[foo.bar]giz.h" (vms)? Are they accepted at
> all (I couldn't find any tests for those for the windows platform)? I'm not
> sure on how one should to represent these in the portable path notation.
I don't know, since I never used such path. I think it is resonable to ban
them, especially considering that no Windows user ever complained.
> Also, I have some "echo" statements that I'd (might) like to keep (but
> normally the should run silent) in the path conversion methods. Is there
> any supported way of doing this so that they are only output when only
> running the build system tests, i.e. some trace level when testing the
> build system?
No, there's no "offiial" way. However, you can use a variable which would be
set to 'true' in the test code for path.jam.
> And, what about using assertions inside the build system code (excluding
> the test code), is it considered ok or does it slow down stuff too much?
I know that there are some asserts in the build system code, for example, in
the generators module, so generally its OK.
Just make sure that each specific assert indicates a programmer's error, not a
user error. The asserts in generators sometimes trigger due to user's error,
which is bad.
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