djones@megatest.UUCP (Dave Jones) (01/10/91)
From article <2865@hsi86.hsi.com>, by may@hsi86.hsi.com (Jeanette May): > We are attempting to compile ANSI C source on 4.3 BSD UNIX (VAX) > environment. I'm wondering why there is seems to be no ANSI-C to C compiler around, a la "cfront", the C++ to C compiler. It would seem to be a natural. There are a couple of things that would have to be kluged up a little -- passing floats rather than doubles as parameters, for example -- but that's no big deal. Is there a market for such a beast? Is there one available? If not, I'm sure I could rig one up without too much trouble.
gwyn@smoke.brl.mil (Doug Gwyn) (01/11/91)
In article <14824@goofy.megatest.UUCP> djones@megatest.UUCP (Dave Jones) writes: >I'm wondering why there is seems to be no ANSI-C to C compiler around, >a la "cfront", the C++ to C compiler. It would seem to be a natural. >There are a couple of things that would have to be kluged up a little -- >passing floats rather than doubles as parameters, for example -- but >that's no big deal. First of all, the ANSI-C to C translator is called "cat" on UNIX systems. Of course, what you really meant was a translator from standard C to some flavor of pre-ANSI C; unfortunately there are numerous incompatible dialects of the latter. Even if you narrow that down to be "4.3BSD PCC's dialect of C", the translation is by no means trivial -- a conforming implementation would be about as difficult to obtain that way as writing a genuine compiler. It is not clear that a non-conforming attempt would be worth all that much.
djones@megatest.UUCP (Dave Jones) (01/11/91)
The topic under consideration is a compiler from ANSI-C to
original C, ("C Classic"?), for those of us who have UNIX boxes
with C, but not the ANSI extensions and changes -- Sun workstations
in my case. I'm seeing more and more stuff on the net in ANSI-C, which
I cannot use easily, and it would be nice to have some of the features
for development work also, particularly prototypes.
So, what exactly are the tricky bits?
I mentioned passing floats (not doubles) to procedures. Ditto chars.
Raymond Chen reminded me of the ANSI rule against rearranging
floating-point expressions.
What else?