RE: hard links vs. sym links

[Gary Oberbrunner]

[Greg mentioned his :S NT extension, and rethinking design goals if they
make things difficult, to see if everything can be reconciled.]

The design goal in question makes sure the compiler finds *generated* header
files in the way you'd expect.

Say a.c includes "gen.h" which is generated by gen.exe.  Cons will run
gen.exe in the build dir to create gen.h in that dir (not in the source
dir).  How should the compiler, compiling a.c in the source dir, then find
gen.h?  Cons could add a -I option, but then it has to figure out how many
levels up and down to go to get to the build or include dir (and there could
be more than one, right?)  This could probably be done though -- it's the
way that would make most sense to me.

An alternative that would work on some systems would be for the compiler to
be run with the build dir as its working dir; then the object files would
have no path info, just simple filenames.  Don't know if that would work on
win32 though.

. . . . . . . . . . . . . . . . . . . . . . . . .
Gary Oberbrunner                address@hidden
GenArts, Inc.                   Tel: 617-492-2888
8 Clinton Street                Fax: 617-492-2852
Cambridge, MA 02139 USA         http://web.genarts.com