Summary [Re: bcopy not in libc.a ?]

From: David M. Meyer 503/346-1747 (
Date: Tue Jul 06 1993 - 22:39:13 CDT

        Well, since I made so many mistakes (possibly starting
        with my posting of my porting notes :-)) and so many good
        responses were given, I though I'd summarize what I've

        (i). bcopy

                I had been using

                #define bcopy(src,dest,len) (memcpy(dest,src,len))
                Now, it was pointed by many out that if src and
                dest overlap, this is broken. One possibility is

                #define bcopy(src,dest,len) (memmove(dest, src, len))

                which according to the man page should handle
                the overlaping case . Another possiblitity is

                #define bcopy(src, dest, cnt) \
                ({ char *cccc; cccc= (char *) \
                 malloc(cnt); memcpy(cccc,src,cnt); \
                 memcpy(dest,cccc,cnt); free(cccc); })

                This was from (Pedro Acebes Bayon).

        (ii). getcwd

                I was missing an argument. Guy Harris and others
                #define getwd(x) getcwd((x), <some number>)

                "where <some number> might be MAXPATHLEN if you've
                included a header file that defines it."

        (iii). random
                I had been using

                #define random rand

                As (Davin Milun) points out,
                "random() and rand() are no where near

                He correctly proposes

                   #define random lrand48

        (iv). Signals

                Casper [ (Casper H.S. Dik)] once
                again was a great source here. As he points out

                "sigblock (if not SVR4 segment) is equivalent to:

                sigaddset(&set, SIGWHAT);
                sigaddset(&set, SIGOTHER);
                sigprocmask(SIG_BLOCK,&set, &oset);
                /* do stuff between blocked signals */
                sigprocmask(SIG_SETMASK,&oset, 0);
                This is the easiers way to install sort of
                BSD-ish compatible signal handlers. The main
                difference between SunOS signal() and SVR4
                sigset() is the restarting of system calls. In
                SunOS (since 4.0 (or was it 4.1?)) system calls
                are restarted. In SVR4 system calls are
                interupted and return EINTR, unless the system
                call is read, write and the like, in which case
                the number of bytes read/written is returned when
                that number exceeds 0. If no bytes have been
                read/written, EINTR is returned. "

        So, I guess we've agreed on the following (add Casper's
        execellent comments):

                #ifdef SOLARIS2
                #include <stdlib.h>
                #define srandom srand
                #define random lrand48
                #define bcopy(src,dest,len) (memmove(dest, src, len))
                /* or possibly
                #define bcopy(src, dest, cnt) ({ char *cccc; \
                cccc= (char *) malloc(cnt); memcpy(cccc,src,cnt); \
                memcpy(dest,cccc,cnt); free(cccc); })
                #define bzero(dest,len) (memset(dest, (char)NULL, len))
                #define bcmp(b1,b2,n) (memcmp(b1,b2,n))
                #define index(s,c) strchr(s,c)
                #define rindex(s,c) strrchr(s,c)
                #define MIN(a, b) ((a) < (b) ? (a) : (b))
                #define MAX(a, b) ((a) < (b) ? (b) : (a))
                #define getwd(x) getcwd((x), <some number>)
                /* where <some number> might be MAXPATHLEN if you've
                included a header file that defines it. */
                #endif /* SOLARIS2 */

                #ifdef SOLARIS2
                    setvbuf(stderr, NULL, _IOLBF, 0);
                #endif /* SOLARIS2 */

                #ifdef SOLARIS2
                        long gethostid() {
                          char buf[128];
                          if (sysinfo(SI_HW_SERIAL, buf, 128) == -1) {
                #endif /* SOLARIS2 */

        Finally, I find notes of this sort really useful. Maybe
        someone should collect them all into a "Solaris 2.x
        porting FAQ". I would be willing to compile such a FAQ if
        people think it would be useful (and one doesn't already
        exist, other than the Answerbook). Let me know.

        At any rate, I really appreciate all the helpful



        David M. Meyer Voice: 503/346-1747
        Senior Network Engineer Pager: 503/342-9458
        Office of University Computing FAX: 503/346-4397
        Computing Center Internet:
        University of Oregon
        1225 Kincaid
        Eugene, OR 97403

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:00 CDT