Bruno Haible <[EMAIL PROTECTED]> wrote: >> 2008-10-18 Bruno Haible <[EMAIL PROTECTED]> >> >> * lib/unistd.in.h (getusershell, setusershell, endusershell): New >> declarations. >> * lib/getusershell.c: Include unistd.h. >> * m4/getusershell.m4 (gl_FUNC_GETUSERSHELL): Require >> gl_UNISTD_H_DEFAULTS and AC_USE_SYSTEM_EXTENSIONS. Set >> HAVE_GETUSERSHELL. >> * m4/unistd_h.m4 (gl_UNISTD_H_DEFAULTS): Initialize GNULIB_GETUSERSHELL >> and HAVE_GETUSERSHELL. >> * modules/getusershell (Depends-on): Add unistd, extensions. >> (configure.ac): Invoke gl_UNISTD_MODULE_INDICATOR. >> (Include): Specify <unistd.h>. >> * modules/unistd (Makefile.am): Substitute GNULIB_GETUSERSHELL and >> HAVE_GETUSERSHELL. > > Applied. > > I now see that coreutils/src/su.c has this: > > > /* Hide any system prototype for getusershell. > This is necessary because some Cray systems have a conflicting > prototype (returning `int') in <unistd.h>. */ > #define getusershell _getusershell_sys_proto_ > > #include "system.h" > #include "getpass.h" > > #undef getusershell > > > If Cray systems are still a portability target, the current idiom for this > kind of problem is to add > #define getusershell rpl_getusershell > in unistd.in.h.
AFAIK, the affected versions of Unicos are no longer in use, so I wouldn't bother. At first I removed those declarations from su.c, but then the ones from the system unistd.h were not visible on recent Linux. Combine that with the fact that coreutils is getting out of the business of maintaining su.c, I opted not to change it at all.
