> ### FROM: Felix <address@hidden>
> ### ON: Fri, 15 Feb 2013 20:49:20 0100 (CET)
>
> From: Vitaly Magerya <address@hidden>
> Subject: Re: [Chicken-users] Error with clang/FreeBSD
> Date: Fri, 15 Feb 2013 20:07:30 0200
>
> > Previously I wrote:
> >> The correct thing to do here is to change '-z origin' into
> >> '-Wl,-z,origin'. This way both GCC and Clang will do the right thing.
> >>
> >> (I currently tested this by compiling C files; I'll be able to test csc
> >> modified that way in a dozen of hours or so, unless any of you folks do
> >> it first).
> >
> > Yup, just tested csc modified that way; -deploy works correctly with
> > gcc, clang and even pcc.
> >
>
> Did you build chicken with pcc? Are there any modifications necessary?
> I would be very interested about the details. I wasn't able to build
> a working pcc so far.
>
>
> cheers,
> felix
>
I tried pcc a couple of years ago, though rather disappointing that compiling chicken with it failed miserably. I don't have the details now, but vaguely remember pcc couldn't handle some gcc-specific requirements as they were at the time.
However using gcc-compiled chicken/csc -c on scheme source, I was able to compile the C files with gcc, pcc and clang and compare the results. Curiously, the clang and pcc apps were the just about the same size, in one example, clang (54402 b), pcc (54414 b), whereas gcc-compiled app was somewhat smaller (38572 b). Notably they all ran just fine, and I didn't notice significant performance differences.
Don't know if anyone has tried other competent unix C compilers, e.g., from Intel. It seems likely there are enough quirky differences among C compilers that it's difficult to create C source code universally compilable even on the same platform. Standards are supposed to account for it, but we all know how that goes.
J. Altfas