Protoize CL and ECL

[olebole]

This is done on an Ubuntu Trusty machine, with the command

protoize -g -l -N -c "-I${iraf}unix/hlib/libc" *.c

There was already an attempt to protoize CL and ECL in the NOIRLab tree (f46b827), but this was not applicable, because the source trees deviate now significantly here because of #32 and a number of other bug fixes and clean-ups that were applied here but not in the NOIRLab tree. Also, it was incomplete, there are quite a number of prototypes missing. However, I checked this commit completely for potential bug fixes to be applied here (no findings for here; but NOIRLab obviously has some bugs...). The joys of missing coordination and a forked source tree.

The protoize tool is unfortunately removed from the gcc distribution after Ubuntu Trusty, so we had to use a VM to finally get this done. It seems that there are no more missing prototypes in IRAF; at least protoize didn't convert any other C file.

Another change here is that in sys/libc, the memory allocation functions should be declared in a more recent way, i.e. the memory is of type void * instead of char *.

This way, we found one bug in pkg/ecl/pfiles.c#L911-L915 (traceable back to version 2.8, 1989!):

    for (d=0;  d < size_arr;  d++) {
	*q = memneed (btoi(len));
	strncpy (*q++, *p++, len-1);
	*(q+len-1) = '\0' ;
    }

(with char **p, **p;)
Obviously, the last line should force-add a '\0' at the end of each string (in case the length exceeded the maximum), but that is somehow different. NOIRLabs change

diff --git a/pkg/ecl/pfiles.c b/pkg/ecl/pfiles.c
index 08de7aa92..585d9954a 100644
--- a/pkg/ecl/pfiles.c
+++ b/pkg/ecl/pfiles.c
@@ -910,7 +948,7 @@ pfilecopy (register struct pfile *pfp)
    for (d=0;  d < size_arr;  d++) {
        *q = memneed (btoi(len));
        strncpy (*q++, *p++, len-1);
-       *(q+len-1) = '\0' ;
+       *(q+len-1) = (char *)'\0' ;
    }

just silences the warning, but doesn't resolve the bug. IMO correct is

    for (d=0;  d < size_arr; d++, q++, p++) {
	*q = memneed (btoi(len));
	strncpy (*q, *p, len-1);
	(*q)[len-1] = '\0';
    }

@mjfitzpatrick, you should change this in your fork as well.

We also take the opportunity here to fix few other (minor) warnings.