From: Toon Knapen (toon.knapen_at_[hidden])
Date: 2003-07-10 07:09:56
> >> I'm also having trouble using the build.sh script. First of all,
> >> 'whence' does not seem to exist on this platform.
> Unfortunately doing `whence 1>/dev/null 2>/dev/null` doesn't
> seem to get rid of the error message in all shells :-( The
> command detects if whence is available, if it fails it uses 'hash'.
Well, I have to comment the body of test_path() out to make it work.
> Now for vacpp, I do try and detect it by seeing if 'xlc' is
> in the path. Is 'xlc' not normally in the path for the
> compiler? Is there a standard place the compiler is
> installed? And like above, what does 'uname' return?
Yes it's in the path. But invoking 'xlc' (without arguments) results in
the man pages being displayed and thus the command never returns unless
hitting 'q' or <space> a few times.
> Those are sometimes in <unistd.h>. It's already included in
> that file (and in other places), so you can try adding
> defined(OS_SUNOS) to the guard.
What guard ? I did'nt see one.
But I've finally succeeded in porting bjam to solaris/sunpro. All
changes are minor including :
_Assignemnt from void* to other type needs cast
_added includes of system headers (but using the pp only done for sun)
_added includes of other headers in jam_src (don't understand why other
compilers don't complain about this)
_related to the previous, I added also some signatures to headers (of
functions defined in the corresponding .c)
(e.g. added prototype to filesys.h)
Another remark I have is that filesys.h provides function signatures for
functions that are not defined in filesys.c.
And finally, I can't link bjam because a definition for
regex_compile(char*) (although the signature is defined in compile.h,
the definition is not in compile.c ?)
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