On Monday 20 January 2003 2:29 am, Steven G. Johnson wrote: > > Right now, Paul Brook has to maintain is own set of > > patches to autoconf[1] and automake to allow us to build our runtime > > library (which is partially written in Fortran 90). > > Thanks, I took a look at the F95 patches for g95...it looks mainly like a > copy of fortran.m4 to f95.m4 with s/f77/f95/. Is that accurate, or could > you point to any more substantive changes you needed to make? My > proposed autoconf patch should do what you want, while sharing the > fortran backends between dialects.
I think that's pretty much it. I made some more substantial changes to automake, to allow dependency tracking of f9x modules. This is overkill for libgfor, which doesn't use modules, but I use it for some of my own programs. This is also the cause of most of the problems as I haven't ironed out all the bugs yet (I think I may also have uncovered a couple of automake bugs). > (I'm a bit surprised that you needed this at all, since I would have > thought you could just use AC_PROG_F77(f90) and compile all Fortran xx > files with $F77, especially since the backend is equivalent anyway in > your implementation. I've done this for existing projects of mine where > I need to compile f90 files, with no major problems. The current > autoconf f77 stuff works for newer dialects for most purposes, except > when one needs to mix f77 and f9x. My patch does address one additional > issue, the fact that IBM's xlf does not accept .f9* extensions without a > special flag, grrr.) I can't remember if I tried this or not. Maybe I'll give it annother try, it'd be nuce if we could do without my hacks. Paul
