todo.irix.dif

minicom的源码,linux下常用的串口程序.

From: "Dominique Avatravaux" <[e-mail]>
Date: Wed, 29 May 1996 15:43:46 -0600
To: [e-mail]
Subject: Minicom patch : porting to IRIX and Solaris

   Hello,

  Here is a patch needed to compile Minicom under IRIX 5.3 and Solaris 2.3/2.4.
The two compilations were done with gcc. The most interesting points are :

   * There is no Makefile modification needed : the Linux entry works fine in
both cases.
   * IRIX has no usleep() library call at all, and some Solaris boxes have
one and others don't, causing the binaries to be incompatible amongst the same
machines :-( (works with libc dated oct 27 1995 and not with the one dated mai
23 1995, no time for further investigation, sorry). ok the source from GNU
libc to fill this hole and put it in sysdep2.c; didn't dare to put it in
sysdep1 since I would have had to name it "m_usleep" :-). The side effect is
that "runscript" now needs sysdep2.o as well.

  Hope this is useful to you ! Please forgive my approximate English, I'm
french (as may *NOT* be seen on my email adress... My real one follows).

  Note : you two (Miquel and Joshua) had better agree on this strange
dual-headed Minicom release. I don't know which kind of relationship exists
between you two and I don't care about it, but please have a thought for your
users : they surely would appreciate a unified version. And a schism is a
pointless workforce dispersion in any way...

  With best regards and great consideration for your work,

--
						Dominique Quatravaux
					[e-mail]
--
diff -r -C 4 minicom-1.74/src/Makefile minicom-1.74.orig/src/Makefile
*** minicom-1.74/src/Makefile	Wed May 29 15:01:20 1996
--- minicom-1.74.orig/src/Makefile	Sun Apr 21 03:41:12 1996
***************
*** 143,151 ****
  	  windiv.o sysdep1.o sysdep2.o rwconf.o main.o

  KOBJS	= keyserv.o wkeys.o sysdep2.o

! SOBJS	= script.o sysdep1.o sysdep2.o

  CFLAGS	= $(FLAGS) -DLIBDIR=\"$(LIBDIR)\"

  R	= $(ROOTDIR)
--- 143,151 ----
  	  windiv.o sysdep1.o sysdep2.o rwconf.o main.o

  KOBJS	= keyserv.o wkeys.o sysdep2.o

! SOBJS	= script.o sysdep1.o

  CFLAGS	= $(FLAGS) -DLIBDIR=\"$(LIBDIR)\"

  R	= $(ROOTDIR)
diff -r -C 4 minicom-1.74/src/port.h minicom-1.74.orig/src/port.h
*** minicom-1.74/src/port.h	Mon May 27 19:22:56 1996
--- minicom-1.74.orig/src/port.h	Thu Jan 18 19:44:48 1996
***************
*** 104,113 ****
  #  if !defined(_SVR2) && !defined(_SYSV3) && !defined(_COH42)
  #    define HAS_FCHOWN
  #  endif
  #endif
-
- /* The usleep existence test needs to be factorized somehow, but the
#inclusion
-  scheme does not allow it right now. */
- #if defined(__sgi__) || ( defined (__sun__) && defined (__svr4__))
-  int usleep(unsigned int usec);
- #endif
--- 104,107 ----
diff -r -C 4 minicom-1.74/src/sysdep.h minicom-1.74.orig/src/sysdep.h
*** minicom-1.74/src/sysdep.h	Wed May 29 14:38:23 1996
--- minicom-1.74.orig/src/sysdep.h	Thu Jan 18 19:44:49 1996
***************
*** 153,164 ****
  #  define cfsetispeed(xtty, xspd) \
  		((xtty)->c_cflag = ((xtty)->c_cflag & ~CBAUD) | (xspd))
  #  define cfsetospeed(tty, spd)
  #endif
-
- /* IRIX has no usleep, nor do some versions of Solaris 2.4's libc.  */
- #if defined(__sgi__) || ( defined (__sun__) && defined (__svr4__))
- #  define NO_USLEEP
- #  include <sys/time.h>
- #  include <signal.h>
-    int usleep(unsigned int useconds);
- #endif
--- 153,156 ----
diff -r -C 4 minicom-1.74/src/sysdep2.c minicom-1.74.orig/src/sysdep2.c
*** minicom-1.74/src/sysdep2.c	Wed May 29 14:38:50 1996
--- minicom-1.74.orig/src/sysdep2.c	Thu Jan 18 19:44:49 1996
***************
*** 267,287 ****
    return(exstat);
  }
  #endif

- /* If we have no usleep, it's about time we make one... Excerpt from
-   glibc 1.09.1.
-   Could stupid OSes lack both select() and usleep() ? I hope not ! */
- #ifdef NO_USLEEP
- int usleep(unsigned int useconds)
- {
-   struct timeval delay;
-
-   delay.tv_sec = 0;
-   delay.tv_usec = useconds;
-
-   (void) select (0, (fd_set *) NULL, (fd_set *) NULL, (fd_set *) NULL,
-                    &delay);
-
-   return 0;
- }
- #endif
--- 267,270 ----


-- 
							Dominique Quatravaux
						[e-mail]