Port eBBS 3.1.4-rc2 to build and run on macOS/Darwin (Apple Silicon)

[DrPitre]

Summary

• Adds a DARWIN os profile in osdeps.h (based on FREEBSD) so the codebase can be identified/built on macOS
• Updates Makefile CCFLAGS for modern clang's stricter K&R diagnostics (implicit-int, implicit-function-declaration are hard errors by default now); omits -lcrypt on Darwin (crypt is in libSystem)
• Fixes terminal I/O on macOS: replaces ioctl(TIOCGETA/TIOCSETA) with tcgetattr()/tcsetattr() in pbbs/term.c — the ioctl path returned EFAULT/Bad address on Apple Silicon
• Fixes two variadic functions (prints() in pbbs/screen.c, bbslog() in log.c) that lacked prototypes visible at their call sites — on ARM64 this caused va_arg() to read corrupted values, resulting in SIGSEGV crashes during login; adds extern declarations to clientui.h and pbbs/io.h so all translation units see the correct variadic ABI
• Converts every K&R-style function definition across the codebase (~54 files) to ANSI C prototype form; zero warnings on modern clang
• Fixes gram.y: converts K&R function definitions to ANSI C, adds forward declarations for all grammar-action functions, removes obsolete inline K&R-style local declarations; regenerates y.tab.c
• Adds the missing standard-library #includes and cross-file function prototypes needed once -Wimplicit-function-declaration (a hard error by default on modern clang) was no longer suppressed
• Fixes the outc/outcf termcap callback signatures to match tputs()'s real prototype on macOS's ncurses-backed termcap.h
• Fixes two buffer-overflow-adjacent strncpy size arguments (login.c terminal field, bbsmaild.c sender name)
• Fixes a handful of genuine call-site bugs uncovered by the now-correct prototypes (missing execute() argument in netmail.c, SelectBoard/yyerror signature mismatches in nmenus.c)
• Adds etc -> config symlink and .gitignore entries for runtime files (log, tmp/, .shmkey, .utable)

Test plan

• make clean && make builds all binaries (lbbs, chatd, addacct, delacct, bbslog, bbfinger, bbsmaild) with 0 errors and 0 warnings on macOS/Apple Silicon with Xcode clang
• Verified full interactive login flow (SYSOP account) reaches the Main Menu without crashing
• Smoke-tested all seven binaries' CLI entry points (usage output / expected startup behavior) after the full ANSI-C conversion

[catskillmarina]

thanks

[virtadpt]

Hi! I saw your PR and I'll get it merged as soon as I can get it to build on my laptop. I don't have a Mac to test on I'm afraid, but I did try to compile it on my Linux box and it failed about halfway through. I saw that you added some more changes and once I get them pulled into my local copy I'll give it another try.

[virtadpt]

make clobber passed. make all failed with an error in login.c:

gcc -O -DCOLOR -std=gnu89    -c -o login.o login.c
login.c: In function ‘get_remote_host’:
login.c:40:21: error: ‘UTMPX_FILE’ undeclared (first use in this function); did you mean ‘UTMP_FILE’?
   40 | #  define UTMP_FILE UTMPX_FILE
      |                     ^~~~~~~~~~
login.c:152:18: note: in expansion of macro ‘UTMP_FILE’
  152 |   if ((fd = open(UTMP_FILE, O_RDONLY)) != -1) {
      |                  ^~~~~~~~~
login.c:40:21: note: each undeclared identifier is reported only once for each function it appears in
login.c:152:18: note: in expansion of macro ‘UTMP_FILE’
  152 |   if ((fd = open(UTMP_FILE, O_RDONLY)) != -1) {
      |                  ^~~~~~~~~
make: *** [<builtin>: login.o] Error 1

I'm not sure yet if this is due to your changes or a bug that needs fixed in general.

[DrPitre]

Thanks for testing on Linux! I just pushed a fix for that. The root cause was that the macOS port hardcoded DARWIN=1 and FREEBSD=1 in osdeps.h, which got picked up on your Linux build and triggered the utmpx path — but Linux's <utmpx.h> doesn't define UTMPX_FILE the way macOS does.

The fix does two things:

1. osdeps.h: Platform flags are now auto-detected from compiler predefined macros (__APPLE__, __FreeBSD__, __linux__), so the file no longer needs to be manually edited per platform.

2. login.c: When UTMPX_FILE isn't defined (Linux), it now falls back to _PATH_UTMPX from <paths.h>, then /var/run/utmpx as a last resort.

make clobber && make all should work on your Linux box now without any manual changes. Let me know how it goes!

[catskillmarina]

It seems the original author is still working on this -> https://wq5l.net/ebbs/ .

I never knew !

[virtadpt]

@DrPitre Thank you - I'll give it a try later today.

@catskillmarina I had no idea - holy crap!

[DrPitre]

I just updated my fork to Ray's 3.1.4 rc2 code, so this is the latest!

[virtadpt]

Because I was having some trouble, I erased my old clone of the repo with your PR applied and re-pulled it. make all still fails:

gcc -O -DCOLOR -Wno-implicit-function-declaration -Wno-implicit-int    -c -o client.o client.c
client.c: In function ‘hit_alarm_clock’:
client.c:139:19: error: passing argument 2 of ‘signal’ from incompatible pointer type [[-Wincompatible-pointer-types](https://gcc.gnu.org/onlinedocs/gcc-16.1.0/gcc/Warning-Options.html#index-Wincompatible-pointer-types)]
  139 |   signal(SIGALRM, hit_alarm_clock);
      |                   ^~~~~~~~~~~~~~~
      |                   |
      |                   void (*)(void)
In file included from client.c:22:
/usr/include/signal.h:88:57: note: expected ‘__sighandler_t’ {aka ‘void (*)(int’} but argument is of type ‘void (*)(void)’
   88 | extern __sighandler_t signal (int __sig, __sighandler_t __handler)
      |                                          ~~~~~~~~~~~~~~~^~~~~~~~~
client.c:133:1: note: ‘hit_alarm_clock’ declared here
  133 | hit_alarm_clock()
      | ^~~~~~~~~~~~~~~
/usr/include/signal.h:72:16: note: ‘__sighandler_t’ declared here
   72 | typedef void (*__sighandler_t) (int);
      |                ^~~~~~~~~~~~~~
client.c: In function ‘set_idle_alarm’:
client.c:146:21: error: passing argument 2 of ‘signal’ from incompatible pointer type [[-Wincompatible-pointer-types](https://gcc.gnu.org/onlinedocs/gcc-16.1.0/gcc/Warning-Options.html#index-Wincompatible-pointer-types)]
  146 |     signal(SIGALRM, hit_alarm_clock);
      |                     ^~~~~~~~~~~~~~~
      |                     |
      |                     void (*)(void)
/usr/include/signal.h:88:57: note: expected ‘__sighandler_t’ {aka ‘void (*)(int’} but argument is of type ‘void (*)(void)’
   88 | extern __sighandler_t signal (int __sig, __sighandler_t __handler)
      |                                          ~~~~~~~~~~~~~~~^~~~~~~~~
client.c:133:1: note: ‘hit_alarm_clock’ declared here
  133 | hit_alarm_clock()
      | ^~~~~~~~~~~~~~~
/usr/include/signal.h:72:16: note: ‘__sighandler_t’ declared here
   72 | typedef void (*__sighandler_t) (int);
      |                ^~~~~~~~~~~~~~
make: *** [<builtin>: client.o] Error 1

[DrPitre]

Thanks for testing again! Fixed — the issue was that hit_alarm_clock() in client.c was declared K&R-style with no parameters, which under strict modern compilers (gcc 16, clang) resolves to void (*)(void) instead of the void (*)(int) that signal() expects.

Pushed a fix (899ca42) that gives it an explicit int parameter. make clobber && make all builds clean here. Let me know how it goes on your box.