ruby/process.c

3844 строки
88 KiB
C

/**********************************************************************
process.c -
$Author$
$Date$
created at: Tue Aug 10 14:30:50 JST 1993
Copyright (C) 1993-2003 Yukihiro Matsumoto
Copyright (C) 2000 Network Applied Communication Laboratory, Inc.
Copyright (C) 2000 Information-technology Promotion Agency, Japan
**********************************************************************/
#include "ruby.h"
#include "rubysig.h"
#include <stdio.h>
#include <errno.h>
#include <signal.h>
#ifdef HAVE_STDLIB_H
#include <stdlib.h>
#endif
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
#ifdef HAVE_FCNTL_H
#include <fcntl.h>
#endif
#ifdef __DJGPP__
#include <process.h>
#endif
#include <time.h>
#include <ctype.h>
#ifndef EXIT_SUCCESS
#define EXIT_SUCCESS 0
#endif
#ifndef EXIT_FAILURE
#define EXIT_FAILURE 1
#endif
struct timeval rb_time_interval _((VALUE));
#ifdef HAVE_SYS_WAIT_H
# include <sys/wait.h>
#endif
#ifdef HAVE_SYS_RESOURCE_H
# include <sys/resource.h>
#endif
#include "st.h"
#ifdef __EMX__
#undef HAVE_GETPGRP
#endif
#ifdef HAVE_SYS_TIMES_H
#include <sys/times.h>
#endif
#ifdef HAVE_GRP_H
#include <grp.h>
#endif
#if defined(HAVE_TIMES) || defined(_WIN32)
static VALUE S_Tms;
#endif
#ifndef WIFEXITED
#define WIFEXITED(w) (((w) & 0xff) == 0)
#endif
#ifndef WIFSIGNALED
#define WIFSIGNALED(w) (((w) & 0x7f) > 0 && (((w) & 0x7f) < 0x7f))
#endif
#ifndef WIFSTOPPED
#define WIFSTOPPED(w) (((w) & 0xff) == 0x7f)
#endif
#ifndef WEXITSTATUS
#define WEXITSTATUS(w) (((w) >> 8) & 0xff)
#endif
#ifndef WTERMSIG
#define WTERMSIG(w) ((w) & 0x7f)
#endif
#ifndef WSTOPSIG
#define WSTOPSIG WEXITSTATUS
#endif
#if defined(__APPLE__) && ( defined(__MACH__) || defined(__DARWIN__) ) && !defined(__MacOS_X__)
#define __MacOS_X__ 1
#endif
#if defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__bsdi__)
#define HAVE_44BSD_SETUID 1
#define HAVE_44BSD_SETGID 1
#endif
#ifdef __NetBSD__
#undef HAVE_SETRUID
#undef HAVE_SETRGID
#endif
#if defined(__MacOS_X__) || defined(__bsdi__)
#define BROKEN_SETREUID 1
#define BROKEN_SETREGID 1
#endif
#if defined(HAVE_44BSD_SETUID) || defined(__MacOS_X__)
#if !defined(USE_SETREUID) && !defined(BROKEN_SETREUID)
#define OBSOLETE_SETREUID 1
#endif
#if !defined(USE_SETREGID) && !defined(BROKEN_SETREGID)
#define OBSOLETE_SETREGID 1
#endif
#endif
#define preserving_errno(stmts) \
do {int saved_errno = errno; stmts; errno = saved_errno;} while (0)
/*
* call-seq:
* Process.pid => fixnum
*
* Returns the process id of this process. Not available on all
* platforms.
*
* Process.pid #=> 27415
*/
static VALUE
get_pid()
{
rb_secure(2);
return INT2FIX(getpid());
}
/*
* call-seq:
* Process.ppid => fixnum
*
* Returns the process id of the parent of this process. Always
* returns 0 on NT. Not available on all platforms.
*
* puts "I am #{Process.pid}"
* Process.fork { puts "Dad is #{Process.ppid}" }
*
* <em>produces:</em>
*
* I am 27417
* Dad is 27417
*/
static VALUE
get_ppid()
{
rb_secure(2);
#ifdef _WIN32
return INT2FIX(0);
#else
return INT2FIX(getppid());
#endif
}
/*********************************************************************
*
* Document-class: Process::Status
*
* <code>Process::Status</code> encapsulates the information on the
* status of a running or terminated system process. The built-in
* variable <code>$?</code> is either +nil+ or a
* <code>Process::Status</code> object.
*
* fork { exit 99 } #=> 26557
* Process.wait #=> 26557
* $?.class #=> Process::Status
* $?.to_i #=> 25344
* $? >> 8 #=> 99
* $?.stopped? #=> false
* $?.exited? #=> true
* $?.exitstatus #=> 99
*
* Posix systems record information on processes using a 16-bit
* integer. The lower bits record the process status (stopped,
* exited, signaled) and the upper bits possibly contain additional
* information (for example the program's return code in the case of
* exited processes). Pre Ruby 1.8, these bits were exposed directly
* to the Ruby program. Ruby now encapsulates these in a
* <code>Process::Status</code> object. To maximize compatibility,
* however, these objects retain a bit-oriented interface. In the
* descriptions that follow, when we talk about the integer value of
* _stat_, we're referring to this 16 bit value.
*/
static VALUE rb_cProcStatus;
VALUE rb_last_status = Qnil;
static void
last_status_set(status, pid)
int status, pid;
{
rb_last_status = rb_obj_alloc(rb_cProcStatus);
rb_iv_set(rb_last_status, "status", INT2FIX(status));
rb_iv_set(rb_last_status, "pid", INT2FIX(pid));
}
/*
* call-seq:
* stat.to_i => fixnum
* stat.to_int => fixnum
*
* Returns the bits in _stat_ as a <code>Fixnum</code>. Poking
* around in these bits is platform dependent.
*
* fork { exit 0xab } #=> 26566
* Process.wait #=> 26566
* sprintf('%04x', $?.to_i) #=> "ab00"
*/
static VALUE
pst_to_i(st)
VALUE st;
{
return rb_iv_get(st, "status");
}
/*
* call-seq:
* stat.to_s => string
*
* Equivalent to _stat_<code>.to_i.to_s</code>.
*/
static VALUE
pst_to_s(st)
VALUE st;
{
return rb_fix2str(pst_to_i(st), 10);
}
/*
* call-seq:
* stat.pid => fixnum
*
* Returns the process ID that this status object represents.
*
* fork { exit } #=> 26569
* Process.wait #=> 26569
* $?.pid #=> 26569
*/
static VALUE
pst_pid(st)
VALUE st;
{
return rb_iv_get(st, "pid");
}
/*
* call-seq:
* stat.inspect => string
*
* Override the inspection method.
*/
static VALUE
pst_inspect(st)
VALUE st;
{
VALUE pid;
int status;
VALUE str;
char buf[256];
pid = pst_pid(st);
status = NUM2INT(st);
snprintf(buf, sizeof(buf), "#<%s: pid=%ld", rb_class2name(CLASS_OF(st)), NUM2LONG(pid));
str = rb_str_new2(buf);
if (WIFSTOPPED(status)) {
int stopsig = WSTOPSIG(status);
const char *signame = ruby_signal_name(stopsig);
if (signame) {
snprintf(buf, sizeof(buf), ",stopped(SIG%s=%d)", signame, stopsig);
}
else {
snprintf(buf, sizeof(buf), ",stopped(%d)", stopsig);
}
rb_str_cat2(str, buf);
}
if (WIFSIGNALED(status)) {
int termsig = WTERMSIG(status);
const char *signame = ruby_signal_name(termsig);
if (signame) {
snprintf(buf, sizeof(buf), ",signaled(SIG%s=%d)", signame, termsig);
}
else {
snprintf(buf, sizeof(buf), ",signaled(%d)", termsig);
}
rb_str_cat2(str, buf);
}
if (WIFEXITED(status)) {
snprintf(buf, sizeof(buf), ",exited(%d)", WEXITSTATUS(status));
rb_str_cat2(str, buf);
}
#ifdef WCOREDUMP
if (WCOREDUMP(status)) {
rb_str_cat2(str, ",coredumped");
}
#endif
rb_str_cat2(str, ">");
return str;
}
/*
* call-seq:
* stat == other => true or false
*
* Returns +true+ if the integer value of _stat_
* equals <em>other</em>.
*/
static VALUE
pst_equal(st1, st2)
VALUE st1, st2;
{
if (st1 == st2) return Qtrue;
return rb_equal(pst_to_i(st1), st2);
}
/*
* call-seq:
* stat & num => fixnum
*
* Logical AND of the bits in _stat_ with <em>num</em>.
*
* fork { exit 0x37 }
* Process.wait
* sprintf('%04x', $?.to_i) #=> "3700"
* sprintf('%04x', $? & 0x1e00) #=> "1600"
*/
static VALUE
pst_bitand(st1, st2)
VALUE st1, st2;
{
int status = NUM2INT(st1) & NUM2INT(st2);
return INT2NUM(status);
}
/*
* call-seq:
* stat >> num => fixnum
*
* Shift the bits in _stat_ right <em>num</em> places.
*
* fork { exit 99 } #=> 26563
* Process.wait #=> 26563
* $?.to_i #=> 25344
* $? >> 8 #=> 99
*/
static VALUE
pst_rshift(st1, st2)
VALUE st1, st2;
{
int status = NUM2INT(st1) >> NUM2INT(st2);
return INT2NUM(status);
}
/*
* call-seq:
* stat.stopped? => true or false
*
* Returns +true+ if this process is stopped. This is only
* returned if the corresponding <code>wait</code> call had the
* <code>WUNTRACED</code> flag set.
*/
static VALUE
pst_wifstopped(st)
VALUE st;
{
int status = NUM2INT(st);
if (WIFSTOPPED(status))
return Qtrue;
else
return Qfalse;
}
/*
* call-seq:
* stat.stopsig => fixnum or nil
*
* Returns the number of the signal that caused _stat_ to stop
* (or +nil+ if self is not stopped).
*/
static VALUE
pst_wstopsig(st)
VALUE st;
{
int status = NUM2INT(st);
if (WIFSTOPPED(status))
return INT2NUM(WSTOPSIG(status));
return Qnil;
}
/*
* call-seq:
* stat.signaled? => true or false
*
* Returns +true+ if _stat_ terminated because of
* an uncaught signal.
*/
static VALUE
pst_wifsignaled(st)
VALUE st;
{
int status = NUM2INT(st);
if (WIFSIGNALED(status))
return Qtrue;
else
return Qfalse;
}
/*
* call-seq:
* stat.termsig => fixnum or nil
*
* Returns the number of the signal that caused _stat_ to
* terminate (or +nil+ if self was not terminated by an
* uncaught signal).
*/
static VALUE
pst_wtermsig(st)
VALUE st;
{
int status = NUM2INT(st);
if (WIFSIGNALED(status))
return INT2NUM(WTERMSIG(status));
return Qnil;
}
/*
* call-seq:
* stat.exited? => true or false
*
* Returns +true+ if _stat_ exited normally (for
* example using an <code>exit()</code> call or finishing the
* program).
*/
static VALUE
pst_wifexited(st)
VALUE st;
{
int status = NUM2INT(st);
if (WIFEXITED(status))
return Qtrue;
else
return Qfalse;
}
/*
* call-seq:
* stat.exitstatus => fixnum or nil
*
* Returns the least significant eight bits of the return code of
* _stat_. Only available if <code>exited?</code> is
* +true+.
*
* fork { } #=> 26572
* Process.wait #=> 26572
* $?.exited? #=> true
* $?.exitstatus #=> 0
*
* fork { exit 99 } #=> 26573
* Process.wait #=> 26573
* $?.exited? #=> true
* $?.exitstatus #=> 99
*/
static VALUE
pst_wexitstatus(st)
VALUE st;
{
int status = NUM2INT(st);
if (WIFEXITED(status))
return INT2NUM(WEXITSTATUS(status));
return Qnil;
}
/*
* call-seq:
* stat.success? => true, false or nil
*
* Returns +true+ if _stat_ is successful, +false+ if not.
* Returns +nil+ if <code>exited?</code> is not +true+.
*/
static VALUE
pst_success_p(st)
VALUE st;
{
int status = NUM2INT(st);
if (!WIFEXITED(status))
return Qnil;
return WEXITSTATUS(status) == EXIT_SUCCESS ? Qtrue : Qfalse;
}
/*
* call-seq:
* stat.coredump? => true or false
*
* Returns +true+ if _stat_ generated a coredump
* when it terminated. Not available on all platforms.
*/
static VALUE
pst_wcoredump(st)
VALUE st;
{
#ifdef WCOREDUMP
int status = NUM2INT(st);
if (WCOREDUMP(status))
return Qtrue;
else
return Qfalse;
#else
return Qfalse;
#endif
}
#if !defined(HAVE_WAITPID) && !defined(HAVE_WAIT4)
#define NO_WAITPID
static st_table *pid_tbl;
#endif
int
rb_waitpid(pid, st, flags)
int pid;
int *st;
int flags;
{
int result;
#ifndef NO_WAITPID
int oflags = flags;
if (!rb_thread_alone()) { /* there're other threads to run */
flags |= WNOHANG;
}
retry:
TRAP_BEG;
#ifdef HAVE_WAITPID
result = waitpid(pid, st, flags);
#else /* HAVE_WAIT4 */
result = wait4(pid, st, flags, NULL);
#endif
TRAP_END;
if (result < 0) {
if (errno == EINTR) {
rb_thread_polling();
goto retry;
}
return -1;
}
if (result == 0) {
if (oflags & WNOHANG) return 0;
rb_thread_polling();
if (rb_thread_alone()) flags = oflags;
goto retry;
}
#else /* NO_WAITPID */
if (pid_tbl && st_lookup(pid_tbl, pid, (st_data_t *)st)) {
last_status_set(*st, pid);
st_delete(pid_tbl, (st_data_t*)&pid, NULL);
return pid;
}
if (flags) {
rb_raise(rb_eArgError, "can't do waitpid with flags");
}
for (;;) {
TRAP_BEG;
result = wait(st);
TRAP_END;
if (result < 0) {
if (errno == EINTR) {
rb_thread_schedule();
continue;
}
return -1;
}
if (result == pid) {
break;
}
if (!pid_tbl)
pid_tbl = st_init_numtable();
st_insert(pid_tbl, pid, (st_data_t)st);
if (!rb_thread_alone()) rb_thread_schedule();
}
#endif
if (result > 0) {
last_status_set(*st, result);
}
return result;
}
#ifdef NO_WAITPID
struct wait_data {
int pid;
int status;
};
static int
wait_each(pid, status, data)
int pid, status;
struct wait_data *data;
{
if (data->status != -1) return ST_STOP;
data->pid = pid;
data->status = status;
return ST_DELETE;
}
static int
waitall_each(pid, status, ary)
int pid, status;
VALUE ary;
{
last_status_set(status, pid);
rb_ary_push(ary, rb_assoc_new(INT2NUM(pid), rb_last_status));
return ST_DELETE;
}
#endif
/* [MG]:FIXME: I wasn't sure how this should be done, since ::wait()
has historically been documented as if it didn't take any arguments
despite the fact that it's just an alias for ::waitpid(). The way I
have it below is more truthful, but a little confusing.
I also took the liberty of putting in the pid values, as they're
pretty useful, and it looked as if the original 'ri' output was
supposed to contain them after "[...]depending on the value of
aPid:".
The 'ansi' and 'bs' formats of the ri output don't display the
definition list for some reason, but the plain text one does.
*/
/*
* call-seq:
* Process.wait() => fixnum
* Process.wait(pid=-1, flags=0) => fixnum
* Process.waitpid(pid=-1, flags=0) => fixnum
*
* Waits for a child process to exit, returns its process id, and
* sets <code>$?</code> to a <code>Process::Status</code> object
* containing information on that process. Which child it waits on
* depends on the value of _pid_:
*
* > 0:: Waits for the child whose process ID equals _pid_.
*
* 0:: Waits for any child whose process group ID equals that of the
* calling process.
*
* -1:: Waits for any child process (the default if no _pid_ is
* given).
*
* < -1:: Waits for any child whose process group ID equals the absolute
* value of _pid_.
*
* The _flags_ argument may be a logical or of the flag values
* <code>Process::WNOHANG</code> (do not block if no child available)
* or <code>Process::WUNTRACED</code> (return stopped children that
* haven't been reported). Not all flags are available on all
* platforms, but a flag value of zero will work on all platforms.
*
* Calling this method raises a <code>SystemError</code> if there are
* no child processes. Not available on all platforms.
*
* include Process
* fork { exit 99 } #=> 27429
* wait #=> 27429
* $?.exitstatus #=> 99
*
* pid = fork { sleep 3 } #=> 27440
* Time.now #=> Wed Apr 09 08:57:09 CDT 2003
* waitpid(pid, Process::WNOHANG) #=> nil
* Time.now #=> Wed Apr 09 08:57:09 CDT 2003
* waitpid(pid, 0) #=> 27440
* Time.now #=> Wed Apr 09 08:57:12 CDT 2003
*/
static VALUE
proc_wait(argc, argv)
int argc;
VALUE *argv;
{
VALUE vpid, vflags;
int pid, flags, status;
rb_secure(2);
flags = 0;
rb_scan_args(argc, argv, "02", &vpid, &vflags);
if (argc == 0) {
pid = -1;
}
else {
pid = NUM2INT(vpid);
if (argc == 2 && !NIL_P(vflags)) {
flags = NUM2UINT(vflags);
}
}
if ((pid = rb_waitpid(pid, &status, flags)) < 0)
rb_sys_fail(0);
if (pid == 0) {
return rb_last_status = Qnil;
}
return INT2FIX(pid);
}
/*
* call-seq:
* Process.wait2(pid=-1, flags=0) => [pid, status]
* Process.waitpid2(pid=-1, flags=0) => [pid, status]
*
* Waits for a child process to exit (see Process::waitpid for exact
* semantics) and returns an array containing the process id and the
* exit status (a <code>Process::Status</code> object) of that
* child. Raises a <code>SystemError</code> if there are no child
* processes.
*
* Process.fork { exit 99 } #=> 27437
* pid, status = Process.wait2
* pid #=> 27437
* status.exitstatus #=> 99
*/
static VALUE
proc_wait2(argc, argv)
int argc;
VALUE *argv;
{
VALUE pid = proc_wait(argc, argv);
if (NIL_P(pid)) return Qnil;
return rb_assoc_new(pid, rb_last_status);
}
/*
* call-seq:
* Process.waitall => [ [pid1,status1], ...]
*
* Waits for all children, returning an array of
* _pid_/_status_ pairs (where _status_ is a
* <code>Process::Status</code> object).
*
* fork { sleep 0.2; exit 2 } #=> 27432
* fork { sleep 0.1; exit 1 } #=> 27433
* fork { exit 0 } #=> 27434
* p Process.waitall
*
* <em>produces</em>:
*
* [[27434, #<Process::Status: pid=27434,exited(0)>],
* [27433, #<Process::Status: pid=27433,exited(1)>],
* [27432, #<Process::Status: pid=27432,exited(2)>]]
*/
static VALUE
proc_waitall()
{
VALUE result;
int pid, status;
rb_secure(2);
result = rb_ary_new();
#ifdef NO_WAITPID
if (pid_tbl) {
st_foreach(pid_tbl, waitall_each, result);
}
for (pid = -1;;) {
pid = wait(&status);
if (pid == -1) {
if (errno == ECHILD)
break;
if (errno == EINTR) {
rb_thread_schedule();
continue;
}
rb_sys_fail(0);
}
last_status_set(status, pid);
rb_ary_push(result, rb_assoc_new(INT2NUM(pid), rb_last_status));
}
#else
rb_last_status = Qnil;
for (pid = -1;;) {
pid = rb_waitpid(-1, &status, 0);
if (pid == -1) {
if (errno == ECHILD)
break;
rb_sys_fail(0);
}
rb_ary_push(result, rb_assoc_new(INT2NUM(pid), rb_last_status));
}
#endif
return result;
}
static VALUE
detach_process_watcher(pid_p)
int *pid_p;
{
int cpid, status;
for (;;) {
cpid = rb_waitpid(*pid_p, &status, WNOHANG);
if (cpid == -1) return rb_last_status;
rb_thread_sleep(1);
}
}
VALUE
rb_detach_process(pid)
int pid;
{
return rb_thread_create(detach_process_watcher, (void*)&pid);
}
/*
* call-seq:
* Process.detach(pid) => thread
*
* Some operating systems retain the status of terminated child
* processes until the parent collects that status (normally using
* some variant of <code>wait()</code>. If the parent never collects
* this status, the child stays around as a <em>zombie</em> process.
* <code>Process::detach</code> prevents this by setting up a
* separate Ruby thread whose sole job is to reap the status of the
* process _pid_ when it terminates. Use <code>detach</code>
* only when you do not intent to explicitly wait for the child to
* terminate. <code>detach</code> only checks the status
* periodically (currently once each second).
*
* The waiting thread returns the exit status of the detached process
* when it terminates, so you can use <code>Thread#join</code> to
* know the result. If specified _pid_ is not a valid child process
* ID, the thread returns +nil+ immediately.
*
* In this first example, we don't reap the first child process, so
* it appears as a zombie in the process status display.
*
* p1 = fork { sleep 0.1 }
* p2 = fork { sleep 0.2 }
* Process.waitpid(p2)
* sleep 2
* system("ps -ho pid,state -p #{p1}")
*
* <em>produces:</em>
*
* 27389 Z
*
* In the next example, <code>Process::detach</code> is used to reap
* the child automatically.
*
* p1 = fork { sleep 0.1 }
* p2 = fork { sleep 0.2 }
* Process.detach(p1)
* Process.waitpid(p2)
* sleep 2
* system("ps -ho pid,state -p #{p1}")
*
* <em>(produces no output)</em>
*/
static VALUE
proc_detach(obj, pid)
VALUE pid;
{
rb_secure(2);
return rb_detach_process(NUM2INT(pid));
}
#ifndef HAVE_STRING_H
char *strtok();
#endif
#ifdef HAVE_SETITIMER
#define before_exec() rb_thread_stop_timer()
#define after_exec() rb_thread_start_timer()
#else
#define before_exec()
#define after_exec()
#endif
extern char *dln_find_exe();
static void
security(str)
const char *str;
{
if (rb_env_path_tainted()) {
if (rb_safe_level() > 0) {
rb_raise(rb_eSecurityError, "Insecure PATH - %s", str);
}
}
}
static int
proc_exec_v(argv, prog)
char **argv;
const char *prog;
{
if (!prog)
prog = argv[0];
security(prog);
prog = dln_find_exe(prog, 0);
if (!prog) {
errno = ENOENT;
return -1;
}
#if (defined(MSDOS) && !defined(DJGPP)) || defined(__human68k__) || defined(__EMX__) || defined(OS2)
{
#if defined(__human68k__)
#define COMMAND "command.x"
#endif
#if defined(__EMX__) || defined(OS2) /* OS/2 emx */
#define COMMAND "cmd.exe"
#endif
#if (defined(MSDOS) && !defined(DJGPP))
#define COMMAND "command.com"
#endif
char *extension;
if ((extension = strrchr(prog, '.')) != NULL && strcasecmp(extension, ".bat") == 0) {
char **new_argv;
char *p;
int n;
for (n = 0; argv[n]; n++)
/* no-op */;
new_argv = ALLOCA_N(char*, n + 2);
for (; n > 0; n--)
new_argv[n + 1] = argv[n];
new_argv[1] = strcpy(ALLOCA_N(char, strlen(argv[0]) + 1), argv[0]);
for (p = new_argv[1]; *p != '\0'; p++)
if (*p == '/')
*p = '\\';
new_argv[0] = COMMAND;
argv = new_argv;
prog = dln_find_exe(argv[0], 0);
if (!prog) {
errno = ENOENT;
return -1;
}
}
}
#endif /* MSDOS or __human68k__ or __EMX__ */
before_exec();
execv(prog, argv);
preserving_errno(after_exec());
return -1;
}
int
rb_proc_exec_n(argc, argv, prog)
int argc;
VALUE *argv;
const char *prog;
{
char **args;
int i;
args = ALLOCA_N(char*, argc+1);
for (i=0; i<argc; i++) {
args[i] = RSTRING(argv[i])->ptr;
}
args[i] = 0;
if (args[0]) {
return proc_exec_v(args, prog);
}
return -1;
}
int
rb_proc_exec(str)
const char *str;
{
const char *s = str;
char *ss, *t;
char **argv, **a;
while (*str && ISSPACE(*str))
str++;
#ifdef _WIN32
before_exec();
rb_w32_spawn(P_OVERLAY, (char *)str, 0);
after_exec();
#else
for (s=str; *s; s++) {
if (ISSPACE(*s)) {
const char *p, *nl = NULL;
for (p = s; ISSPACE(*p); p++) {
if (*p == '\n') nl = p;
}
if (!*p) break;
if (nl) s = nl;
}
if (*s != ' ' && !ISALPHA(*s) && strchr("*?{}[]<>()~&|\\$;'`\"\n",*s)) {
int status;
#if defined(MSDOS)
before_exec();
status = system(str);
after_exec();
if (status != -1)
exit(status);
#elif defined(__human68k__) || defined(__CYGWIN32__) || defined(__EMX__)
char *shell = dln_find_exe("sh", 0);
status = -1;
before_exec();
if (shell)
execl(shell, "sh", "-c", str, (char *) NULL);
else
status = system(str);
after_exec();
if (status != -1)
exit(status);
#else
before_exec();
execl("/bin/sh", "sh", "-c", str, (char *)NULL);
preserving_errno(after_exec());
#endif
return -1;
}
}
a = argv = ALLOCA_N(char*, (s-str)/2+2);
ss = ALLOCA_N(char, s-str+1);
memcpy(ss, str, s-str);
ss[s-str] = '\0';
if (*a++ = strtok(ss, " \t")) {
while (t = strtok(NULL, " \t")) {
*a++ = t;
}
*a = NULL;
}
if (argv[0]) {
return proc_exec_v(argv, 0);
}
errno = ENOENT;
#endif /* _WIN32 */
return -1;
}
#if defined(_WIN32)
#define HAVE_SPAWNV 1
#endif
#if !defined(HAVE_FORK) && defined(HAVE_SPAWNV)
static int
proc_spawn_v(argv, prog)
char **argv;
char *prog;
{
#if defined(_WIN32)
char *cmd = ALLOCA_N(char, rb_w32_argv_size(argv));
if (!prog) prog = argv[0];
return rb_w32_spawn(P_NOWAIT, rb_w32_join_argv(cmd, argv), prog);
#else
char *extension;
int status;
if (!prog)
prog = argv[0];
security(prog);
prog = dln_find_exe(prog, 0);
if (!prog)
return -1;
#if defined(__human68k__)
if ((extension = strrchr(prog, '.')) != NULL && strcasecmp(extension, ".bat") == 0) {
char **new_argv;
char *p;
int n;
for (n = 0; argv[n]; n++)
/* no-op */;
new_argv = ALLOCA_N(char*, n + 2);
for (; n > 0; n--)
new_argv[n + 1] = argv[n];
new_argv[1] = strcpy(ALLOCA_N(char, strlen(argv[0]) + 1), argv[0]);
for (p = new_argv[1]; *p != '\0'; p++)
if (*p == '/')
*p = '\\';
new_argv[0] = COMMAND;
argv = new_argv;
prog = dln_find_exe(argv[0], 0);
if (!prog) {
errno = ENOENT;
return -1;
}
}
#endif
before_exec();
status = spawnv(P_WAIT, prog, argv);
last_status_set(status == -1 ? 127 : status, 0);
after_exec();
return status;
#endif
}
static int
proc_spawn_n(argc, argv, prog)
int argc;
VALUE *argv;
VALUE prog;
{
char **args;
int i;
args = ALLOCA_N(char*, argc + 1);
for (i = 0; i < argc; i++) {
args[i] = RSTRING(argv[i])->ptr;
}
args[i] = (char*) 0;
if (args[0])
return proc_spawn_v(args, prog ? RSTRING(prog)->ptr : 0);
return -1;
}
#if defined(_WIN32)
#define proc_spawn(str) rb_w32_spawn(P_NOWAIT, str, 0)
#else
static int
proc_spawn(str)
char *str;
{
char *s, *t;
char **argv, **a;
int status;
for (s = str; *s; s++) {
if (*s != ' ' && !ISALPHA(*s) && strchr("*?{}[]<>()~&|\\$;'`\"\n",*s)) {
char *shell = dln_find_exe("sh", 0);
before_exec();
status = shell?spawnl(P_WAIT,shell,"sh","-c",str,(char*)NULL):system(str);
last_status_set(status == -1 ? 127 : status, 0);
after_exec();
return status;
}
}
a = argv = ALLOCA_N(char*, (s - str) / 2 + 2);
s = ALLOCA_N(char, s - str + 1);
strcpy(s, str);
if (*a++ = strtok(s, " \t")) {
while (t = strtok(NULL, " \t"))
*a++ = t;
*a = NULL;
}
return argv[0] ? proc_spawn_v(argv, 0) : -1;
}
#endif
#endif
VALUE
rb_check_argv(argc, argv)
int argc;
VALUE *argv;
{
VALUE tmp, prog;
int i;
if (argc == 0) {
rb_raise(rb_eArgError, "wrong number of arguments");
}
prog = 0;
tmp = rb_check_array_type(argv[0]);
if (!NIL_P(tmp)) {
if (RARRAY(tmp)->len != 2) {
rb_raise(rb_eArgError, "wrong first argument");
}
prog = RARRAY(tmp)->ptr[0];
argv[0] = RARRAY(tmp)->ptr[1];
SafeStringValue(prog);
}
for (i = 0; i < argc; i++) {
SafeStringValue(argv[i]);
}
security(RSTRING(prog ? prog : argv[0])->ptr);
return prog;
}
/*
* call-seq:
* exec(command [, arg, ...])
*
* Replaces the current process by running the given external _command_.
* If +exec+ is given a single argument, that argument is
* taken as a line that is subject to shell expansion before being
* executed. If multiple arguments are given, the second and subsequent
* arguments are passed as parameters to _command_ with no shell
* expansion. If the first argument is a two-element array, the first
* element is the command to be executed, and the second argument is
* used as the <code>argv[0]</code> value, which may show up in process
* listings. In MSDOS environments, the command is executed in a
* subshell; otherwise, one of the <code>exec(2)</code> system calls is
* used, so the running command may inherit some of the environment of
* the original program (including open file descriptors).
*
* Raises SystemCallError if the _command_ couldn't execute (typically
* <code>Errno::ENOENT</code> when it was not found).
*
* exec "echo *" # echoes list of files in current directory
* # never get here
*
*
* exec "echo", "*" # echoes an asterisk
* # never get here
*/
VALUE
rb_f_exec(argc, argv)
int argc;
VALUE *argv;
{
struct rb_exec_arg e;
VALUE prog;
prog = rb_check_argv(argc, argv);
if (!prog && argc == 1) {
e.argc = 0;
e.argv = 0;
e.prog = RSTRING(argv[0])->ptr;
}
else {
e.argc = argc;
e.argv = argv;
e.prog = prog ? RSTRING(prog)->ptr : 0;
}
rb_exec(&e);
rb_sys_fail(e.prog);
return Qnil; /* dummy */
}
int
rb_exec(e)
const struct rb_exec_arg *e;
{
int argc = e->argc;
VALUE *argv = e->argv;
const char *prog = e->prog;
if (argc == 0) {
rb_proc_exec(prog);
}
else {
rb_proc_exec_n(argc, argv, prog);
}
#ifndef FD_CLOEXEC
preserving_errno({
fprintf(stderr, "%s:%d: command not found: %s\n",
ruby_sourcefile, ruby_sourceline, prog);
});
#endif
return -1;
}
#ifdef HAVE_FORK
#ifdef FD_CLOEXEC
#if SIZEOF_INT == SIZEOF_LONG
#define proc_syswait (VALUE (*)_((VALUE)))rb_syswait
#else
static VALUE
proc_syswait(pid)
VALUE pid;
{
rb_syswait((int)pid);
return Qnil;
}
#endif
#endif
/*
* Forks child process, and returns the process ID in the parent
* process.
*
* If +status+ is given, protects from any exceptions and sets the
* jump status to it.
*
* In the child process, just returns 0 if +chfunc+ is +NULL+.
* Otherwise +chfunc+ will be called with +charg+, and then the child
* process exits with +EXIT_SUCCESS+ when it returned zero.
*
* In the case of the function is called and returns non-zero value,
* the child process exits with non-+EXIT_SUCCESS+ value (normaly
* 127). And, on the platforms where +FD_CLOEXEC+ is available,
* +errno+ is propagated to the parent process, and this function
* returns -1 in the parent process. On the other platforms, just
* returns pid.
*
* +chfunc+ must not raise any exceptions.
*/
int
rb_fork(status, chfunc, charg)
int *status;
int (*chfunc) _((void *));
void *charg;
{
int pid, err, state = 0;
#ifdef FD_CLOEXEC
int ep[2];
#endif
#ifndef __VMS
rb_io_flush(rb_stdout);
rb_io_flush(rb_stderr);
#endif
#ifdef FD_CLOEXEC
if (chfunc) {
if (pipe(ep)) return -1;
if (fcntl(ep[1], F_SETFD, FD_CLOEXEC)) {
preserving_errno((close(ep[0]), close(ep[1])));
return -1;
}
}
#endif
while ((pid = fork()) < 0) {
switch (errno) {
case EAGAIN:
#if defined(EWOULDBLOCK) && EWOULDBLOCK != EAGAIN
case EWOULDBLOCK:
#endif
if (!status && !chfunc) {
rb_thread_sleep(1);
continue;
}
else {
rb_protect((VALUE (*)())rb_thread_sleep, 1, &state);
if (status) *status = state;
if (!state) continue;
}
default:
#ifdef FD_CLOEXEC
if (chfunc) {
preserving_errno((close(ep[0]), close(ep[1])));
}
#endif
if (state && !status) rb_jump_tag(state);
return -1;
}
}
if (!pid) {
if (chfunc) {
#ifdef FD_CLOEXEC
close(ep[0]);
#endif
if (!(*chfunc)(charg)) _exit(EXIT_SUCCESS);
#ifdef FD_CLOEXEC
err = errno;
write(ep[1], &err, sizeof(err));
#endif
#if EXIT_SUCCESS == 127
_exit(EXIT_FAILURE);
#else
_exit(127);
#endif
}
}
#ifdef FD_CLOEXEC
else if (chfunc) {
close(ep[1]);
if ((state = read(ep[0], &err, sizeof(err))) < 0) {
err = errno;
}
close(ep[0]);
if (state) {
if (status) {
rb_protect(proc_syswait, (VALUE)pid, status);
}
else {
rb_syswait(pid);
}
errno = err;
return -1;
}
}
#endif
return pid;
}
#endif
/*
* call-seq:
* Kernel.fork [{ block }] => fixnum or nil
* Process.fork [{ block }] => fixnum or nil
*
* Creates a subprocess. If a block is specified, that block is run
* in the subprocess, and the subprocess terminates with a status of
* zero. Otherwise, the +fork+ call returns twice, once in
* the parent, returning the process ID of the child, and once in
* the child, returning _nil_. The child process can exit using
* <code>Kernel.exit!</code> to avoid running any
* <code>at_exit</code> functions. The parent process should
* use <code>Process.wait</code> to collect the termination statuses
* of its children or use <code>Process.detach</code> to register
* disinterest in their status; otherwise, the operating system
* may accumulate zombie processes.
*/
static VALUE
rb_f_fork(obj)
VALUE obj;
{
#ifdef HAVE_FORK
int pid;
rb_secure(2);
switch (pid = rb_fork(0, 0, 0)) {
case 0:
#ifdef linux
after_exec();
#endif
rb_thread_atfork();
if (rb_block_given_p()) {
int status;
rb_protect(rb_yield, Qundef, &status);
ruby_stop(status);
}
return Qnil;
case -1:
rb_sys_fail("fork(2)");
return Qnil;
default:
return INT2FIX(pid);
}
#else
rb_notimplement();
#endif
}
/*
* call-seq:
* Process.exit!(fixnum=-1)
*
* Exits the process immediately. No exit handlers are
* run. <em>fixnum</em> is returned to the underlying system as the
* exit status.
*
* Process.exit!(0)
*/
static VALUE
rb_f_exit_bang(argc, argv, obj)
int argc;
VALUE *argv;
VALUE obj;
{
VALUE status;
int istatus;
rb_secure(4);
if (rb_scan_args(argc, argv, "01", &status) == 1) {
switch (status) {
case Qtrue:
istatus = EXIT_SUCCESS;
break;
case Qfalse:
istatus = EXIT_FAILURE;
break;
default:
istatus = NUM2INT(status);
break;
}
}
else {
istatus = EXIT_FAILURE;
}
_exit(istatus);
return Qnil; /* not reached */
}
#if defined(sun)
#define signal(a,b) sigset(a,b)
#endif
void
rb_syswait(pid)
int pid;
{
static int overriding;
RETSIGTYPE (*hfunc)_((int)), (*qfunc)_((int)), (*ifunc)_((int));
int status;
int i, hooked = Qfalse;
if (!overriding) {
#ifdef SIGHUP
hfunc = signal(SIGHUP, SIG_IGN);
#endif
#ifdef SIGQUIT
qfunc = signal(SIGQUIT, SIG_IGN);
#endif
ifunc = signal(SIGINT, SIG_IGN);
overriding = Qtrue;
hooked = Qtrue;
}
do {
i = rb_waitpid(pid, &status, 0);
} while (i == -1 && errno == EINTR);
if (hooked) {
#ifdef SIGHUP
signal(SIGHUP, hfunc);
#endif
#ifdef SIGQUIT
signal(SIGQUIT, qfunc);
#endif
signal(SIGINT, ifunc);
overriding = Qfalse;
}
}
int
rb_spawn(argc, argv)
int argc;
VALUE *argv;
{
int status;
VALUE prog;
#if defined HAVE_FORK
struct rb_exec_arg earg;
#endif
prog = rb_check_argv(argc, argv);
if (!prog && argc == 1) {
--argc;
prog = *argv++;
}
#if defined HAVE_FORK
earg.argc = argc;
earg.argv = argv;
earg.prog = prog ? RSTRING(prog)->ptr : 0;
status = rb_fork(&status, (int (*)_((void*)))rb_exec, &earg);
if (prog && argc) argv[0] = prog;
#elif defined HAVE_SPAWNV
if (!argc) {
status = proc_spawn(RSTRING(prog)->ptr);
}
else {
status = proc_spawn_n(argc, argv, prog);
}
if (prog && argc) argv[0] = prog;
#else
if (prog && argc) argv[0] = prog;
if (argc) prog = rb_ary_join(rb_ary_new4(argc, argv), rb_str_new2(" "));
status = system(StringValuePtr(prog));
# if defined(__human68k__) || defined(__DJGPP__)
last_status_set(status == -1 ? 127 : status, 0);
# else
last_status_set((status & 0xff) << 8, 0);
# endif
#endif
return status;
}
/*
* call-seq:
* system(cmd [, arg, ...]) => true or false
*
* Executes _cmd_ in a subshell, returning +true+ if the command ran
* successfully, +false+ otherwise. An error status is available in
* <code>$?</code>. The arguments are processed in the same way as
* for <code>Kernel::exec</code>, and raises same exceptions as it.
*
* system("echo *")
* system("echo", "*")
*
* <em>produces:</em>
*
* config.h main.rb
* *
*/
static VALUE
rb_f_system(argc, argv)
int argc;
VALUE *argv;
{
int status;
status = rb_spawn(argc, argv);
if (status == -1) rb_sys_fail(RSTRING(argv[0])->ptr);
#if defined(HAVE_FORK) || defined(HAVE_SPAWNV)
rb_syswait(status);
status = NUM2INT(rb_last_status);
#endif
if (status == EXIT_SUCCESS) return Qtrue;
return Qfalse;
}
/*
* call-seq:
* spawn(cmd [, arg, ...]) => pid
*
* Similar to <code>Kernel::system</code> except for not waiting for
* end of _cmd_, but returns its <i>pid</i>.
*/
static VALUE
rb_f_spawn(argc, argv)
int argc;
VALUE *argv;
{
int pid;
pid = rb_spawn(argc, argv);
if (pid == -1) rb_sys_fail(RSTRING(argv[0])->ptr);
#if defined(HAVE_FORK) || defined(HAVE_SPAWNV)
return INT2NUM(pid);
#else
return Qnil;
#endif
}
/*
* call-seq:
* sleep([duration]) => fixnum
*
* Suspends the current thread for _duration_ seconds (which may be
* any number, including a +Float+ with fractional seconds). Returns the actual
* number of seconds slept (rounded), which may be less than that asked
* for if the thread was interrupted by a +SIGALRM+, or if
* another thread calls <code>Thread#run</code>. Zero arguments
* causes +sleep+ to sleep forever.
*
* Time.new #=> Wed Apr 09 08:56:32 CDT 2003
* sleep 1.2 #=> 1
* Time.new #=> Wed Apr 09 08:56:33 CDT 2003
* sleep 1.9 #=> 2
* Time.new #=> Wed Apr 09 08:56:35 CDT 2003
*/
static VALUE
rb_f_sleep(argc, argv)
int argc;
VALUE *argv;
{
int beg, end;
beg = time(0);
if (argc == 0) {
rb_thread_sleep_forever();
}
else if (argc == 1) {
rb_thread_wait_for(rb_time_interval(argv[0]));
}
else {
rb_raise(rb_eArgError, "wrong number of arguments");
}
end = time(0) - beg;
return INT2FIX(end);
}
/*
* call-seq:
* Process.getpgrp => integer
*
* Returns the process group ID for this process. Not available on
* all platforms.
*
* Process.getpgid(0) #=> 25527
* Process.getpgrp #=> 25527
*/
static VALUE
proc_getpgrp()
{
int pgrp;
rb_secure(2);
#if defined(HAVE_GETPGRP) && defined(GETPGRP_VOID)
pgrp = getpgrp();
if (pgrp < 0) rb_sys_fail(0);
return INT2FIX(pgrp);
#else
# ifdef HAVE_GETPGID
pgrp = getpgid(0);
if (pgrp < 0) rb_sys_fail(0);
return INT2FIX(pgrp);
# else
rb_notimplement();
# endif
#endif
}
/*
* call-seq:
* Process.setpgrp => 0
*
* Equivalent to <code>setpgid(0,0)</code>. Not available on all
* platforms.
*/
static VALUE
proc_setpgrp()
{
rb_secure(2);
/* check for posix setpgid() first; this matches the posix */
/* getpgrp() above. It appears that configure will set SETPGRP_VOID */
/* even though setpgrp(0,0) would be prefered. The posix call avoids */
/* this confusion. */
#ifdef HAVE_SETPGID
if (setpgid(0,0) < 0) rb_sys_fail(0);
#elif defined(HAVE_SETPGRP) && defined(SETPGRP_VOID)
if (setpgrp() < 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return INT2FIX(0);
}
/*
* call-seq:
* Process.getpgid(pid) => integer
*
* Returns the process group ID for the given process id. Not
* available on all platforms.
*
* Process.getpgid(Process.ppid()) #=> 25527
*/
static VALUE
proc_getpgid(obj, pid)
VALUE obj, pid;
{
#if defined(HAVE_GETPGID) && !defined(__CHECKER__)
int i;
rb_secure(2);
i = getpgid(NUM2INT(pid));
if (i < 0) rb_sys_fail(0);
return INT2NUM(i);
#else
rb_notimplement();
#endif
}
/*
* call-seq:
* Process.setpgid(pid, integer) => 0
*
* Sets the process group ID of _pid_ (0 indicates this
* process) to <em>integer</em>. Not available on all platforms.
*/
static VALUE
proc_setpgid(obj, pid, pgrp)
VALUE obj, pid, pgrp;
{
#ifdef HAVE_SETPGID
int ipid, ipgrp;
rb_secure(2);
ipid = NUM2INT(pid);
ipgrp = NUM2INT(pgrp);
if (setpgid(ipid, ipgrp) < 0) rb_sys_fail(0);
return INT2FIX(0);
#else
rb_notimplement();
#endif
}
/*
* call-seq:
* Process.setsid => fixnum
*
* Establishes this process as a new session and process group
* leader, with no controlling tty. Returns the session id. Not
* available on all platforms.
*
* Process.setsid #=> 27422
*/
static VALUE
proc_setsid()
{
#if defined(HAVE_SETSID)
int pid;
rb_secure(2);
pid = setsid();
if (pid < 0) rb_sys_fail(0);
return INT2FIX(pid);
#elif defined(HAVE_SETPGRP) && defined(TIOCNOTTY)
pid_t pid;
int ret;
rb_secure(2);
pid = getpid();
#if defined(SETPGRP_VOID)
ret = setpgrp();
/* If `pid_t setpgrp(void)' is equivalent to setsid(),
`ret' will be the same value as `pid', and following open() will fail.
In Linux, `int setpgrp(void)' is equivalent to setpgid(0, 0). */
#else
ret = setpgrp(0, pid);
#endif
if (ret == -1) rb_sys_fail(0);
if ((fd = open("/dev/tty", O_RDWR)) >= 0) {
ioctl(fd, TIOCNOTTY, NULL);
close(fd);
}
return INT2FIX(pid);
#else
rb_notimplement();
#endif
}
/*
* call-seq:
* Process.getpriority(kind, integer) => fixnum
*
* Gets the scheduling priority for specified process, process group,
* or user. <em>kind</em> indicates the kind of entity to find: one
* of <code>Process::PRIO_PGRP</code>,
* <code>Process::PRIO_USER</code>, or
* <code>Process::PRIO_PROCESS</code>. _integer_ is an id
* indicating the particular process, process group, or user (an id
* of 0 means _current_). Lower priorities are more favorable
* for scheduling. Not available on all platforms.
*
* Process.getpriority(Process::PRIO_USER, 0) #=> 19
* Process.getpriority(Process::PRIO_PROCESS, 0) #=> 19
*/
static VALUE
proc_getpriority(obj, which, who)
VALUE obj, which, who;
{
#ifdef HAVE_GETPRIORITY
int prio, iwhich, iwho;
rb_secure(2);
iwhich = NUM2INT(which);
iwho = NUM2INT(who);
errno = 0;
prio = getpriority(iwhich, iwho);
if (errno) rb_sys_fail(0);
return INT2FIX(prio);
#else
rb_notimplement();
#endif
}
/*
* call-seq:
* Process.setpriority(kind, integer, priority) => 0
*
* See <code>Process#getpriority</code>.
*
* Process.setpriority(Process::PRIO_USER, 0, 19) #=> 0
* Process.setpriority(Process::PRIO_PROCESS, 0, 19) #=> 0
* Process.getpriority(Process::PRIO_USER, 0) #=> 19
* Process.getpriority(Process::PRIO_PROCESS, 0) #=> 19
*/
static VALUE
proc_setpriority(obj, which, who, prio)
VALUE obj, which, who, prio;
{
#ifdef HAVE_GETPRIORITY
int iwhich, iwho, iprio;
rb_secure(2);
iwhich = NUM2INT(which);
iwho = NUM2INT(who);
iprio = NUM2INT(prio);
if (setpriority(iwhich, iwho, iprio) < 0)
rb_sys_fail(0);
return INT2FIX(0);
#else
rb_notimplement();
#endif
}
#if SIZEOF_RLIM_T == SIZEOF_INT
# define RLIM2NUM(v) UINT2NUM(v)
# define NUM2RLIM(v) NUM2UINT(v)
#elif SIZEOF_RLIM_T == SIZEOF_LONG
# define RLIM2NUM(v) ULONG2NUM(v)
# define NUM2RLIM(v) NUM2ULONG(v)
#elif SIZEOF_RLIM_T == SIZEOF_LONG_LONG
# define RLIM2NUM(v) ULL2NUM(v)
# define NUM2RLIM(v) NUM2ULL(v)
#endif
/*
* call-seq:
* Process.getrlimit(resource) => [cur_limit, max_limit]
*
* Gets the resource limit of the process.
* _cur_limit_ means current (soft) limit and
* _max_limit_ means maximum (hard) limit.
*
* _resource_ indicates the kind of resource to limit:
* such as <code>Process::RLIMIT_CORE</code>,
* <code>Process::RLIMIT_CPU</code>, etc.
* See Process.setrlimit for details.
*
* _cur_limit_ and _max_limit_ may be <code>Process::RLIM_INFINITY</code>,
* <code>Process::RLIM_SAVED_MAX</code> or
* <code>Process::RLIM_SAVED_CUR</code>.
* See Process.setrlimit and the system getrlimit(2) manual for details.
*/
static VALUE
proc_getrlimit(VALUE obj, VALUE resource)
{
#if defined(HAVE_GETRLIMIT) && defined(RLIM2NUM)
struct rlimit rlim;
rb_secure(2);
if (getrlimit(NUM2INT(resource), &rlim) < 0) {
rb_sys_fail("getrlimit");
}
return rb_assoc_new(RLIM2NUM(rlim.rlim_cur), RLIM2NUM(rlim.rlim_max));
#else
rb_notimplement();
#endif
}
/*
* call-seq:
* Process.setrlimit(resource, cur_limit, max_limit) => nil
*
* Sets the resource limit of the process.
* _cur_limit_ means current (soft) limit and
* _max_limit_ means maximum (hard) limit.
*
* _resource_ indicates the kind of resource to limit.
* Although the list of resources are OS dependent,
* SUSv3 defines following resources.
*
* [Process::RLIMIT_CORE] core size (bytes)
* [Process::RLIMIT_CPU] CPU time (seconds)
* [Process::RLIMIT_DATA] data segment (bytes)
* [Process::RLIMIT_FSIZE] file size (bytes)
* [Process::RLIMIT_NOFILE] file descriptors (number)
* [Process::RLIMIT_STACK] stack size (bytes)
* [Process::RLIMIT_AS] total available memory (bytes)
*
* Other <code>Process::RLIMIT_???</code> constants may be defined.
*
* _cur_limit_ and _max_limit_ may be <code>Process::RLIM_INFINITY</code>,
* which means that the resource is not limited.
* They may be <code>Process::RLIM_SAVED_MAX</code> or
* <code>Process::RLIM_SAVED_CUR</code> too.
* See system setrlimit(2) manual for details.
*
*/
static VALUE
proc_setrlimit(VALUE obj, VALUE resource, VALUE rlim_cur, VALUE rlim_max)
{
#if defined(HAVE_SETRLIMIT) && defined(NUM2RLIM)
struct rlimit rlim;
rb_secure(2);
rlim.rlim_cur = NUM2RLIM(rlim_cur);
rlim.rlim_max = NUM2RLIM(rlim_max);
if (setrlimit(NUM2INT(resource), &rlim) < 0) {
rb_sys_fail("setrlimit");
}
return Qnil;
#else
rb_notimplement();
#endif
}
static int under_uid_switch = 0;
static void
check_uid_switch()
{
rb_secure(2);
if (under_uid_switch) {
rb_raise(rb_eRuntimeError, "can't handle UID while evaluating block given to Process::UID.switch method");
}
}
static int under_gid_switch = 0;
static void
check_gid_switch()
{
rb_secure(2);
if (under_gid_switch) {
rb_raise(rb_eRuntimeError, "can't handle GID while evaluating block given to Process::UID.switch method");
}
}
/*********************************************************************
* Document-class: Process::Sys
*
* The <code>Process::Sys</code> module contains UID and GID
* functions which provide direct bindings to the system calls of the
* same names instead of the more-portable versions of the same
* functionality found in the <code>Process</code>,
* <code>Process::UID</code>, and <code>Process::GID</code> modules.
*/
/*
* call-seq:
* Process::Sys.setuid(integer) => nil
*
* Set the user ID of the current process to _integer_. Not
* available on all platforms.
*
*/
static VALUE
p_sys_setuid(obj, id)
VALUE obj, id;
{
#if defined HAVE_SETUID
check_uid_switch();
if (setuid(NUM2INT(id)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.setruid(integer) => nil
*
* Set the real user ID of the calling process to _integer_.
* Not available on all platforms.
*
*/
static VALUE
p_sys_setruid(obj, id)
VALUE obj, id;
{
#if defined HAVE_SETRUID
check_uid_switch();
if (setruid(NUM2INT(id)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.seteuid(integer) => nil
*
* Set the effective user ID of the calling process to
* _integer_. Not available on all platforms.
*
*/
static VALUE
p_sys_seteuid(obj, id)
VALUE obj, id;
{
#if defined HAVE_SETEUID
check_uid_switch();
if (seteuid(NUM2INT(id)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.setreuid(rid, eid) => nil
*
* Sets the (integer) real and/or effective user IDs of the current
* process to _rid_ and _eid_, respectively. A value of
* <code>-1</code> for either means to leave that ID unchanged. Not
* available on all platforms.
*
*/
static VALUE
p_sys_setreuid(obj, rid, eid)
VALUE obj, rid, eid;
{
#if defined HAVE_SETREUID
check_uid_switch();
if (setreuid(NUM2INT(rid),NUM2INT(eid)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.setresuid(rid, eid, sid) => nil
*
* Sets the (integer) real, effective, and saved user IDs of the
* current process to _rid_, _eid_, and _sid_ respectively. A
* value of <code>-1</code> for any value means to
* leave that ID unchanged. Not available on all platforms.
*
*/
static VALUE
p_sys_setresuid(obj, rid, eid, sid)
VALUE obj, rid, eid, sid;
{
#if defined HAVE_SETRESUID
check_uid_switch();
if (setresuid(NUM2INT(rid),NUM2INT(eid),NUM2INT(sid)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process.uid => fixnum
* Process::UID.rid => fixnum
* Process::Sys.getuid => fixnum
*
* Returns the (real) user ID of this process.
*
* Process.uid #=> 501
*/
static VALUE
proc_getuid(obj)
VALUE obj;
{
int uid = getuid();
return INT2FIX(uid);
}
/*
* call-seq:
* Process.uid= integer => numeric
*
* Sets the (integer) user ID for this process. Not available on all
* platforms.
*/
static VALUE
proc_setuid(obj, id)
VALUE obj, id;
{
int uid = NUM2INT(id);
check_uid_switch();
#if defined(HAVE_SETRESUID) && !defined(__CHECKER__)
if (setresuid(uid, -1, -1) < 0) rb_sys_fail(0);
#elif defined HAVE_SETREUID
if (setreuid(uid, -1) < 0) rb_sys_fail(0);
#elif defined HAVE_SETRUID
if (setruid(uid) < 0) rb_sys_fail(0);
#elif defined HAVE_SETUID
{
if (geteuid() == uid) {
if (setuid(uid) < 0) rb_sys_fail(0);
}
else {
rb_notimplement();
}
}
#else
rb_notimplement();
#endif
return INT2FIX(uid);
}
/********************************************************************
*
* Document-class: Process::UID
*
* The <code>Process::UID</code> module contains a collection of
* module functions which can be used to portably get, set, and
* switch the current process's real, effective, and saved user IDs.
*
*/
static int SAVED_USER_ID;
/*
* call-seq:
* Process::UID.change_privilege(integer) => fixnum
*
* Change the current process's real and effective user ID to that
* specified by _integer_. Returns the new user ID. Not
* available on all platforms.
*
* [Process.uid, Process.euid] #=> [0, 0]
* Process::UID.change_privilege(31) #=> 31
* [Process.uid, Process.euid] #=> [31, 31]
*/
static VALUE
p_uid_change_privilege(obj, id)
VALUE obj, id;
{
extern int errno;
int uid;
check_uid_switch();
uid = NUM2INT(id);
if (geteuid() == 0) { /* root-user */
#if defined(HAVE_SETRESUID)
if (setresuid(uid, uid, uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
#elif defined(HAVE_SETUID)
if (setuid(uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
#elif defined(HAVE_SETREUID) && !defined(OBSOLETE_SETREUID)
if (getuid() == uid) {
if (SAVED_USER_ID == uid) {
if (setreuid(-1, uid) < 0) rb_sys_fail(0);
} else {
if (uid == 0) { /* (r,e,s) == (root, root, x) */
if (setreuid(-1, SAVED_USER_ID) < 0) rb_sys_fail(0);
if (setreuid(SAVED_USER_ID, 0) < 0) rb_sys_fail(0);
SAVED_USER_ID = 0; /* (r,e,s) == (x, root, root) */
if (setreuid(uid, uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
} else {
if (setreuid(0, -1) < 0) rb_sys_fail(0);
SAVED_USER_ID = 0;
if (setreuid(uid, uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
}
}
} else {
if (setreuid(uid, uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
}
#elif defined(HAVE_SETRUID) && defined(HAVE_SETEUID)
if (getuid() == uid) {
if (SAVED_USER_ID == uid) {
if (seteuid(uid) < 0) rb_sys_fail(0);
} else {
if (uid == 0) {
if (setruid(SAVED_USER_ID) < 0) rb_sys_fail(0);
SAVED_USER_ID = 0;
if (setruid(0) < 0) rb_sys_fail(0);
} else {
if (setruid(0) < 0) rb_sys_fail(0);
SAVED_USER_ID = 0;
if (seteuid(uid) < 0) rb_sys_fail(0);
if (setruid(uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
}
}
} else {
if (seteuid(uid) < 0) rb_sys_fail(0);
if (setruid(uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
}
#else
rb_notimplement();
#endif
} else { /* unprivileged user */
#if defined(HAVE_SETRESUID)
if (setresuid((getuid() == uid)? -1: uid,
(geteuid() == uid)? -1: uid,
(SAVED_USER_ID == uid)? -1: uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
#elif defined(HAVE_SETREUID) && !defined(OBSOLETE_SETREUID)
if (SAVED_USER_ID == uid) {
if (setreuid((getuid() == uid)? -1: uid,
(geteuid() == uid)? -1: uid) < 0) rb_sys_fail(0);
} else if (getuid() != uid) {
if (setreuid(uid, (geteuid() == uid)? -1: uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
} else if (/* getuid() == uid && */ geteuid() != uid) {
if (setreuid(geteuid(), uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
if (setreuid(uid, -1) < 0) rb_sys_fail(0);
} else { /* getuid() == uid && geteuid() == uid */
if (setreuid(-1, SAVED_USER_ID) < 0) rb_sys_fail(0);
if (setreuid(SAVED_USER_ID, uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
if (setreuid(uid, -1) < 0) rb_sys_fail(0);
}
#elif defined(HAVE_SETRUID) && defined(HAVE_SETEUID)
if (SAVED_USER_ID == uid) {
if (geteuid() != uid && seteuid(uid) < 0) rb_sys_fail(0);
if (getuid() != uid && setruid(uid) < 0) rb_sys_fail(0);
} else if (/* SAVED_USER_ID != uid && */ geteuid() == uid) {
if (getuid() != uid) {
if (setruid(uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
} else {
if (setruid(SAVED_USER_ID) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
if (setruid(uid) < 0) rb_sys_fail(0);
}
} else if (/* geteuid() != uid && */ getuid() == uid) {
if (seteuid(uid) < 0) rb_sys_fail(0);
if (setruid(SAVED_USER_ID) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
if (setruid(uid) < 0) rb_sys_fail(0);
} else {
errno = EPERM;
rb_sys_fail(0);
}
#elif defined HAVE_44BSD_SETUID
if (getuid() == uid) {
/* (r,e,s)==(uid,?,?) ==> (uid,uid,uid) */
if (setuid(uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
} else {
errno = EPERM;
rb_sys_fail(0);
}
#elif defined HAVE_SETEUID
if (getuid() == uid && SAVED_USER_ID == uid) {
if (seteuid(uid) < 0) rb_sys_fail(0);
} else {
errno = EPERM;
rb_sys_fail(0);
}
#elif defined HAVE_SETUID
if (getuid() == uid && SAVED_USER_ID == uid) {
if (setuid(uid) < 0) rb_sys_fail(0);
} else {
errno = EPERM;
rb_sys_fail(0);
}
#else
rb_notimplement();
#endif
}
return INT2FIX(uid);
}
/*
* call-seq:
* Process::Sys.setgid(integer) => nil
*
* Set the group ID of the current process to _integer_. Not
* available on all platforms.
*
*/
static VALUE
p_sys_setgid(obj, id)
VALUE obj, id;
{
#if defined HAVE_SETGID
check_gid_switch();
if (setgid(NUM2INT(id)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.setrgid(integer) => nil
*
* Set the real group ID of the calling process to _integer_.
* Not available on all platforms.
*
*/
static VALUE
p_sys_setrgid(obj, id)
VALUE obj, id;
{
#if defined HAVE_SETRGID
check_gid_switch();
if (setrgid(NUM2INT(id)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.setegid(integer) => nil
*
* Set the effective group ID of the calling process to
* _integer_. Not available on all platforms.
*
*/
static VALUE
p_sys_setegid(obj, id)
VALUE obj, id;
{
#if defined HAVE_SETEGID
check_gid_switch();
if (setegid(NUM2INT(id)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.setregid(rid, eid) => nil
*
* Sets the (integer) real and/or effective group IDs of the current
* process to <em>rid</em> and <em>eid</em>, respectively. A value of
* <code>-1</code> for either means to leave that ID unchanged. Not
* available on all platforms.
*
*/
static VALUE
p_sys_setregid(obj, rid, eid)
VALUE obj, rid, eid;
{
#if defined HAVE_SETREGID
check_gid_switch();
if (setregid(NUM2INT(rid),NUM2INT(eid)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.setresgid(rid, eid, sid) => nil
*
* Sets the (integer) real, effective, and saved user IDs of the
* current process to <em>rid</em>, <em>eid</em>, and <em>sid</em>
* respectively. A value of <code>-1</code> for any value means to
* leave that ID unchanged. Not available on all platforms.
*
*/
static VALUE
p_sys_setresgid(obj, rid, eid, sid)
VALUE obj, rid, eid, sid;
{
#if defined HAVE_SETRESGID
check_gid_switch();
if (setresgid(NUM2INT(rid),NUM2INT(eid),NUM2INT(sid)) != 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return Qnil;
}
/*
* call-seq:
* Process::Sys.issetugid => true or false
*
* Returns +true+ if the process was created as a result
* of an execve(2) system call which had either of the setuid or
* setgid bits set (and extra privileges were given as a result) or
* if it has changed any of its real, effective or saved user or
* group IDs since it began execution.
*
*/
static VALUE
p_sys_issetugid(obj)
VALUE obj;
{
#if defined HAVE_ISSETUGID
rb_secure(2);
if (issetugid()) {
return Qtrue;
} else {
return Qfalse;
}
#else
rb_notimplement();
return Qnil; /* not reached */
#endif
}
/*
* call-seq:
* Process.gid => fixnum
* Process::GID.rid => fixnum
* Process::Sys.getgid => fixnum
*
* Returns the (real) group ID for this process.
*
* Process.gid #=> 500
*/
static VALUE
proc_getgid(obj)
VALUE obj;
{
int gid = getgid();
return INT2FIX(gid);
}
/*
* call-seq:
* Process.gid= fixnum => fixnum
*
* Sets the group ID for this process.
*/
static VALUE
proc_setgid(obj, id)
VALUE obj, id;
{
int gid = NUM2INT(id);
check_gid_switch();
#if defined(HAVE_SETRESGID) && !defined(__CHECKER__)
if (setresgid(gid, -1, -1) < 0) rb_sys_fail(0);
#elif defined HAVE_SETREGID
if (setregid(gid, -1) < 0) rb_sys_fail(0);
#elif defined HAVE_SETRGID
if (setrgid((GIDTYPE)gid) < 0) rb_sys_fail(0);
#elif defined HAVE_SETGID
{
if (getegid() == gid) {
if (setgid(gid) < 0) rb_sys_fail(0);
}
else {
rb_notimplement();
}
}
#else
rb_notimplement();
#endif
return INT2FIX(gid);
}
static size_t maxgroups = 32;
/*
* call-seq:
* Process.groups => array
*
* Get an <code>Array</code> of the gids of groups in the
* supplemental group access list for this process.
*
* Process.groups #=> [27, 6, 10, 11]
*
*/
static VALUE
proc_getgroups(VALUE obj)
{
#ifdef HAVE_GETGROUPS
VALUE ary;
size_t ngroups;
gid_t *groups;
int i;
groups = ALLOCA_N(gid_t, maxgroups);
ngroups = getgroups(maxgroups, groups);
if (ngroups == -1)
rb_sys_fail(0);
ary = rb_ary_new();
for (i = 0; i < ngroups; i++)
rb_ary_push(ary, INT2NUM(groups[i]));
return ary;
#else
rb_notimplement();
return Qnil;
#endif
}
/*
* call-seq:
* Process.groups= array => array
*
* Set the supplemental group access list to the given
* <code>Array</code> of group IDs.
*
* Process.groups #=> [0, 1, 2, 3, 4, 6, 10, 11, 20, 26, 27]
* Process.groups = [27, 6, 10, 11] #=> [27, 6, 10, 11]
* Process.groups #=> [27, 6, 10, 11]
*
*/
static VALUE
proc_setgroups(VALUE obj, VALUE ary)
{
#ifdef HAVE_SETGROUPS
size_t ngroups;
gid_t *groups;
int i;
struct group *gr;
Check_Type(ary, T_ARRAY);
ngroups = RARRAY(ary)->len;
if (ngroups > maxgroups)
rb_raise(rb_eArgError, "too many groups, %d max", maxgroups);
groups = ALLOCA_N(gid_t, ngroups);
for (i = 0; i < ngroups && i < RARRAY(ary)->len; i++) {
VALUE g = RARRAY(ary)->ptr[i];
if (FIXNUM_P(g)) {
groups[i] = FIX2INT(g);
}
else {
VALUE tmp = rb_check_string_type(g);
if (NIL_P(tmp)) {
groups[i] = NUM2INT(g);
}
else {
gr = getgrnam(RSTRING(tmp)->ptr);
if (gr == NULL)
rb_raise(rb_eArgError,
"can't find group for %s", RSTRING(tmp)->ptr);
groups[i] = gr->gr_gid;
}
}
}
i = setgroups(ngroups, groups);
if (i == -1)
rb_sys_fail(0);
return proc_getgroups(obj);
#else
rb_notimplement();
return Qnil;
#endif
}
/*
* call-seq:
* Process.initgroups(username, gid) => array
*
* Initializes the supplemental group access list by reading the
* system group database and using all groups of which the given user
* is a member. The group with the specified <em>gid</em> is also
* added to the list. Returns the resulting <code>Array</code> of the
* gids of all the groups in the supplementary group access list. Not
* available on all platforms.
*
* Process.groups #=> [0, 1, 2, 3, 4, 6, 10, 11, 20, 26, 27]
* Process.initgroups( "mgranger", 30 ) #=> [30, 6, 10, 11]
* Process.groups #=> [30, 6, 10, 11]
*
*/
static VALUE
proc_initgroups(obj, uname, base_grp)
VALUE obj, uname, base_grp;
{
#ifdef HAVE_INITGROUPS
if (initgroups(StringValuePtr(uname), (gid_t)NUM2INT(base_grp)) != 0) {
rb_sys_fail(0);
}
return proc_getgroups(obj);
#else
rb_notimplement();
return Qnil;
#endif
}
/*
* call-seq:
* Process.maxgroups => fixnum
*
* Returns the maximum number of gids allowed in the supplemental
* group access list.
*
* Process.maxgroups #=> 32
*/
static VALUE
proc_getmaxgroups(obj)
VALUE obj;
{
return INT2FIX(maxgroups);
}
/*
* call-seq:
* Process.maxgroups= fixnum => fixnum
*
* Sets the maximum number of gids allowed in the supplemental group
* access list.
*/
static VALUE
proc_setmaxgroups(obj, val)
VALUE obj;
{
size_t ngroups = FIX2INT(val);
if (ngroups > 4096)
ngroups = 4096;
maxgroups = ngroups;
return INT2FIX(maxgroups);
}
/*
* call-seq:
* Process.daemon() => fixnum
* Process.daemon(nochdir=nil,noclose=nil) => fixnum
*
* Detach the process from controlling terminal and run in
* the background as system daemon. Unless the argument
* nochdir is true (i.e. non false), it changes the current
* working directory to the root ("/"). Unless the argument
* noclose is true, daemon() will redirect standard input,
* standard output and standard error to /dev/null.
*/
static VALUE
proc_daemon(argc, argv)
int argc;
VALUE *argv;
{
VALUE nochdir, noclose;
int n;
rb_scan_args(argc, argv, "02", &nochdir, &noclose);
#if defined(HAVE_DAEMON)
n = daemon(RTEST(nochdir), RTEST(noclose));
if (n < 0) rb_sys_fail("daemon");
return INT2FIX(n);
#elif defined(HAVE_FORK)
switch (rb_fork(0, 0, 0)) {
case -1:
return (-1);
case 0:
break;
default:
_exit(0);
}
proc_setsid();
if (!RTEST(nochdir))
(void)chdir("/");
if (!RTEST(noclose) && (n = open("/dev/null", O_RDWR, 0)) != -1) {
(void)dup2(n, 0);
(void)dup2(n, 1);
(void)dup2(n, 2);
if (n > 2)
(void)close (n);
}
return INT2FIX(0);
#else
rb_notimplement();
#endif
}
/********************************************************************
*
* Document-class: Process::GID
*
* The <code>Process::GID</code> module contains a collection of
* module functions which can be used to portably get, set, and
* switch the current process's real, effective, and saved group IDs.
*
*/
static int SAVED_GROUP_ID;
/*
* call-seq:
* Process::GID.change_privilege(integer) => fixnum
*
* Change the current process's real and effective group ID to that
* specified by _integer_. Returns the new group ID. Not
* available on all platforms.
*
* [Process.gid, Process.egid] #=> [0, 0]
* Process::GID.change_privilege(33) #=> 33
* [Process.gid, Process.egid] #=> [33, 33]
*/
static VALUE
p_gid_change_privilege(obj, id)
VALUE obj, id;
{
extern int errno;
int gid;
check_gid_switch();
gid = NUM2INT(id);
if (geteuid() == 0) { /* root-user */
#if defined(HAVE_SETRESGID)
if (setresgid(gid, gid, gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
#elif defined HAVE_SETGID
if (setgid(gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
#elif defined(HAVE_SETREGID) && !defined(OBSOLETE_SETREGID)
if (getgid() == gid) {
if (SAVED_GROUP_ID == gid) {
if (setregid(-1, gid) < 0) rb_sys_fail(0);
} else {
if (gid == 0) { /* (r,e,s) == (root, y, x) */
if (setregid(-1, SAVED_GROUP_ID) < 0) rb_sys_fail(0);
if (setregid(SAVED_GROUP_ID, 0) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = 0; /* (r,e,s) == (x, root, root) */
if (setregid(gid, gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
} else { /* (r,e,s) == (z, y, x) */
if (setregid(0, 0) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = 0;
if (setregid(gid, gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
}
}
} else {
if (setregid(gid, gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
}
#elif defined(HAVE_SETRGID) && defined (HAVE_SETEGID)
if (getgid() == gid) {
if (SAVED_GROUP_ID == gid) {
if (setegid(gid) < 0) rb_sys_fail(0);
} else {
if (gid == 0) {
if (setegid(gid) < 0) rb_sys_fail(0);
if (setrgid(SAVED_GROUP_ID) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = 0;
if (setrgid(0) < 0) rb_sys_fail(0);
} else {
if (setrgid(0) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = 0;
if (setegid(gid) < 0) rb_sys_fail(0);
if (setrgid(gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
}
}
} else {
if (setegid(gid) < 0) rb_sys_fail(0);
if (setrgid(gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
}
#else
rb_notimplement();
#endif
} else { /* unprivileged user */
#if defined(HAVE_SETRESGID)
if (setresgid((getgid() == gid)? -1: gid,
(getegid() == gid)? -1: gid,
(SAVED_GROUP_ID == gid)? -1: gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
#elif defined(HAVE_SETREGID) && !defined(OBSOLETE_SETREGID)
if (SAVED_GROUP_ID == gid) {
if (setregid((getgid() == gid)? -1: gid,
(getegid() == gid)? -1: gid) < 0) rb_sys_fail(0);
} else if (getgid() != gid) {
if (setregid(gid, (getegid() == gid)? -1: gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
} else if (/* getgid() == gid && */ getegid() != gid) {
if (setregid(getegid(), gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
if (setregid(gid, -1) < 0) rb_sys_fail(0);
} else { /* getgid() == gid && getegid() == gid */
if (setregid(-1, SAVED_GROUP_ID) < 0) rb_sys_fail(0);
if (setregid(SAVED_GROUP_ID, gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
if (setregid(gid, -1) < 0) rb_sys_fail(0);
}
#elif defined(HAVE_SETRGID) && defined(HAVE_SETEGID)
if (SAVED_GROUP_ID == gid) {
if (getegid() != gid && setegid(gid) < 0) rb_sys_fail(0);
if (getgid() != gid && setrgid(gid) < 0) rb_sys_fail(0);
} else if (/* SAVED_GROUP_ID != gid && */ getegid() == gid) {
if (getgid() != gid) {
if (setrgid(gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
} else {
if (setrgid(SAVED_GROUP_ID) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
if (setrgid(gid) < 0) rb_sys_fail(0);
}
} else if (/* getegid() != gid && */ getgid() == gid) {
if (setegid(gid) < 0) rb_sys_fail(0);
if (setrgid(SAVED_GROUP_ID) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
if (setrgid(gid) < 0) rb_sys_fail(0);
} else {
errno = EPERM;
rb_sys_fail(0);
}
#elif defined HAVE_44BSD_SETGID
if (getgid() == gid) {
/* (r,e,s)==(gid,?,?) ==> (gid,gid,gid) */
if (setgid(gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
} else {
errno = EPERM;
rb_sys_fail(0);
}
#elif defined HAVE_SETEGID
if (getgid() == gid && SAVED_GROUP_ID == gid) {
if (setegid(gid) < 0) rb_sys_fail(0);
} else {
errno = EPERM;
rb_sys_fail(0);
}
#elif defined HAVE_SETGID
if (getgid() == gid && SAVED_GROUP_ID == gid) {
if (setgid(gid) < 0) rb_sys_fail(0);
} else {
errno = EPERM;
rb_sys_fail(0);
}
#else
rb_notimplement();
#endif
}
return INT2FIX(gid);
}
/*
* call-seq:
* Process.euid => fixnum
* Process::UID.eid => fixnum
* Process::Sys.geteuid => fixnum
*
* Returns the effective user ID for this process.
*
* Process.euid #=> 501
*/
static VALUE
proc_geteuid(obj)
VALUE obj;
{
int euid = geteuid();
return INT2FIX(euid);
}
/*
* call-seq:
* Process.euid= integer
*
* Sets the effective user ID for this process. Not available on all
* platforms.
*/
static VALUE
proc_seteuid(obj, euid)
VALUE obj, euid;
{
check_uid_switch();
#if defined(HAVE_SETRESUID) && !defined(__CHECKER__)
if (setresuid(-1, NUM2INT(euid), -1) < 0) rb_sys_fail(0);
#elif defined HAVE_SETREUID
if (setreuid(-1, NUM2INT(euid)) < 0) rb_sys_fail(0);
#elif defined HAVE_SETEUID
if (seteuid(NUM2INT(euid)) < 0) rb_sys_fail(0);
#elif defined HAVE_SETUID
euid = NUM2INT(euid);
if (euid == getuid()) {
if (setuid(euid) < 0) rb_sys_fail(0);
}
else {
rb_notimplement();
}
#else
rb_notimplement();
#endif
return euid;
}
static VALUE
rb_seteuid_core(euid)
int euid;
{
int uid;
check_uid_switch();
uid = getuid();
#if defined(HAVE_SETRESUID) && !defined(__CHECKER__)
if (uid != euid) {
if (setresuid(-1,euid,euid) < 0) rb_sys_fail(0);
SAVED_USER_ID = euid;
} else {
if (setresuid(-1,euid,-1) < 0) rb_sys_fail(0);
}
#elif defined(HAVE_SETREUID) && !defined(OBSOLETE_SETREUID)
if (setreuid(-1, euid) < 0) rb_sys_fail(0);
if (uid != euid) {
if (setreuid(euid,uid) < 0) rb_sys_fail(0);
if (setreuid(uid,euid) < 0) rb_sys_fail(0);
SAVED_USER_ID = euid;
}
#elif defined HAVE_SETEUID
if (seteuid(euid) < 0) rb_sys_fail(0);
#elif defined HAVE_SETUID
if (geteuid() == 0) rb_sys_fail(0);
if (setuid(euid) < 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return INT2FIX(euid);
}
/*
* call-seq:
* Process::UID.grant_privilege(integer) => fixnum
* Process::UID.eid= integer => fixnum
*
* Set the effective user ID, and if possible, the saved user ID of
* the process to the given _integer_. Returns the new
* effective user ID. Not available on all platforms.
*
* [Process.uid, Process.euid] #=> [0, 0]
* Process::UID.grant_privilege(31) #=> 31
* [Process.uid, Process.euid] #=> [0, 31]
*/
static VALUE
p_uid_grant_privilege(obj, id)
VALUE obj, id;
{
return rb_seteuid_core(NUM2INT(id));
}
/*
* call-seq:
* Process.egid => fixnum
* Process::GID.eid => fixnum
* Process::Sys.geteid => fixnum
*
* Returns the effective group ID for this process. Not available on
* all platforms.
*
* Process.egid #=> 500
*/
static VALUE
proc_getegid(obj)
VALUE obj;
{
int egid = getegid();
return INT2FIX(egid);
}
/*
* call-seq:
* Process.egid = fixnum => fixnum
*
* Sets the effective group ID for this process. Not available on all
* platforms.
*/
static VALUE
proc_setegid(obj, egid)
VALUE obj, egid;
{
check_gid_switch();
#if defined(HAVE_SETRESGID) && !defined(__CHECKER__)
if (setresgid(-1, NUM2INT(egid), -1) < 0) rb_sys_fail(0);
#elif defined HAVE_SETREGID
if (setregid(-1, NUM2INT(egid)) < 0) rb_sys_fail(0);
#elif defined HAVE_SETEGID
if (setegid(NUM2INT(egid)) < 0) rb_sys_fail(0);
#elif defined HAVE_SETGID
egid = NUM2INT(egid);
if (egid == getgid()) {
if (setgid(egid) < 0) rb_sys_fail(0);
}
else {
rb_notimplement();
}
#else
rb_notimplement();
#endif
return egid;
}
static VALUE
rb_setegid_core(egid)
int egid;
{
int gid;
check_gid_switch();
gid = getgid();
#if defined(HAVE_SETRESGID) && !defined(__CHECKER__)
if (gid != egid) {
if (setresgid(-1,egid,egid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = egid;
} else {
if (setresgid(-1,egid,-1) < 0) rb_sys_fail(0);
}
#elif defined(HAVE_SETREGID) && !defined(OBSOLETE_SETREGID)
if (setregid(-1, egid) < 0) rb_sys_fail(0);
if (gid != egid) {
if (setregid(egid,gid) < 0) rb_sys_fail(0);
if (setregid(gid,egid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = egid;
}
#elif defined HAVE_SETEGID
if (setegid(egid) < 0) rb_sys_fail(0);
#elif defined HAVE_SETGID
if (geteuid() == 0 /* root user */) rb_sys_fail(0);
if (setgid(egid) < 0) rb_sys_fail(0);
#else
rb_notimplement();
#endif
return INT2FIX(egid);
}
/*
* call-seq:
* Process::GID.grant_privilege(integer) => fixnum
* Process::GID.eid = integer => fixnum
*
* Set the effective group ID, and if possible, the saved group ID of
* the process to the given _integer_. Returns the new
* effective group ID. Not available on all platforms.
*
* [Process.gid, Process.egid] #=> [0, 0]
* Process::GID.grant_privilege(31) #=> 33
* [Process.gid, Process.egid] #=> [0, 33]
*/
static VALUE
p_gid_grant_privilege(obj, id)
VALUE obj, id;
{
return rb_setegid_core(NUM2INT(id));
}
/*
* call-seq:
* Process::UID.re_exchangeable? => true or false
*
* Returns +true+ if the real and effective user IDs of a
* process may be exchanged on the current platform.
*
*/
static VALUE
p_uid_exchangeable()
{
#if defined(HAVE_SETRESUID) && !defined(__CHECKER__)
return Qtrue;
#elif defined(HAVE_SETREUID) && !defined(OBSOLETE_SETREUID)
return Qtrue;
#else
return Qfalse;
#endif
}
/*
* call-seq:
* Process::UID.re_exchange => fixnum
*
* Exchange real and effective user IDs and return the new effective
* user ID. Not available on all platforms.
*
* [Process.uid, Process.euid] #=> [0, 31]
* Process::UID.re_exchange #=> 0
* [Process.uid, Process.euid] #=> [31, 0]
*/
static VALUE
p_uid_exchange(obj)
VALUE obj;
{
int uid, euid;
check_uid_switch();
uid = getuid();
euid = geteuid();
#if defined(HAVE_SETRESUID) && !defined(__CHECKER__)
if (setresuid(euid, uid, uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
#elif defined(HAVE_SETREUID) && !defined(OBSOLETE_SETREUID)
if (setreuid(euid,uid) < 0) rb_sys_fail(0);
SAVED_USER_ID = uid;
#else
rb_notimplement();
#endif
return INT2FIX(uid);
}
/*
* call-seq:
* Process::GID.re_exchangeable? => true or false
*
* Returns +true+ if the real and effective group IDs of a
* process may be exchanged on the current platform.
*
*/
static VALUE
p_gid_exchangeable()
{
#if defined(HAVE_SETRESGID) && !defined(__CHECKER__)
return Qtrue;
#elif defined(HAVE_SETREGID) && !defined(OBSOLETE_SETREGID)
return Qtrue;
#else
return Qfalse;
#endif
}
/*
* call-seq:
* Process::GID.re_exchange => fixnum
*
* Exchange real and effective group IDs and return the new effective
* group ID. Not available on all platforms.
*
* [Process.gid, Process.egid] #=> [0, 33]
* Process::GID.re_exchange #=> 0
* [Process.gid, Process.egid] #=> [33, 0]
*/
static VALUE
p_gid_exchange(obj)
VALUE obj;
{
int gid, egid;
check_gid_switch();
gid = getgid();
egid = getegid();
#if defined(HAVE_SETRESGID) && !defined(__CHECKER__)
if (setresgid(egid, gid, gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
#elif defined(HAVE_SETREGID) && !defined(OBSOLETE_SETREGID)
if (setregid(egid,gid) < 0) rb_sys_fail(0);
SAVED_GROUP_ID = gid;
#else
rb_notimplement();
#endif
return INT2FIX(gid);
}
/* [MG] :FIXME: Is this correct? I'm not sure how to phrase this. */
/*
* call-seq:
* Process::UID.sid_available? => true or false
*
* Returns +true+ if the current platform has saved user
* ID functionality.
*
*/
static VALUE
p_uid_have_saved_id()
{
#if defined(HAVE_SETRESUID) || defined(HAVE_SETEUID) || defined(_POSIX_SAVED_IDS)
return Qtrue;
#else
return Qfalse;
#endif
}
#if defined(HAVE_SETRESUID) || defined(HAVE_SETEUID) || defined(_POSIX_SAVED_IDS)
static VALUE
p_uid_sw_ensure(id)
int id;
{
under_uid_switch = 0;
return rb_seteuid_core(id);
}
/*
* call-seq:
* Process::UID.switch => fixnum
* Process::UID.switch {|| block} => object
*
* Switch the effective and real user IDs of the current process. If
* a <em>block</em> is given, the user IDs will be switched back
* after the block is executed. Returns the new effective user ID if
* called without a block, and the return value of the block if one
* is given.
*
*/
static VALUE
p_uid_switch(obj)
VALUE obj;
{
extern int errno;
int uid, euid;
check_uid_switch();
uid = getuid();
euid = geteuid();
if (uid != euid) {
proc_seteuid(obj, INT2FIX(uid));
if (rb_block_given_p()) {
under_uid_switch = 1;
return rb_ensure(rb_yield, Qnil, p_uid_sw_ensure, SAVED_USER_ID);
} else {
return INT2FIX(euid);
}
} else if (euid != SAVED_USER_ID) {
proc_seteuid(obj, INT2FIX(SAVED_USER_ID));
if (rb_block_given_p()) {
under_uid_switch = 1;
return rb_ensure(rb_yield, Qnil, p_uid_sw_ensure, euid);
} else {
return INT2FIX(uid);
}
} else {
errno = EPERM;
rb_sys_fail(0);
}
#else
static VALUE
p_uid_sw_ensure(obj)
VALUE obj;
{
under_uid_switch = 0;
return p_uid_exchange(obj);
}
static VALUE
p_uid_switch(obj)
VALUE obj;
{
extern int errno;
int uid, euid;
check_uid_switch();
uid = getuid();
euid = geteuid();
if (uid == euid) {
errno = EPERM;
rb_sys_fail(0);
}
p_uid_exchange(obj);
if (rb_block_given_p()) {
under_uid_switch = 1;
return rb_ensure(rb_yield, Qnil, p_uid_sw_ensure, obj);
} else {
return INT2FIX(euid);
}
#endif
}
/* [MG] :FIXME: Is this correct? I'm not sure how to phrase this. */
/*
* call-seq:
* Process::GID.sid_available? => true or false
*
* Returns +true+ if the current platform has saved group
* ID functionality.
*
*/
static VALUE
p_gid_have_saved_id()
{
#if defined(HAVE_SETRESGID) || defined(HAVE_SETEGID) || defined(_POSIX_SAVED_IDS)
return Qtrue;
#else
return Qfalse;
#endif
}
#if defined(HAVE_SETRESGID) || defined(HAVE_SETEGID) || defined(_POSIX_SAVED_IDS)
static VALUE
p_gid_sw_ensure(id)
int id;
{
under_gid_switch = 0;
return rb_setegid_core(id);
}
/*
* call-seq:
* Process::GID.switch => fixnum
* Process::GID.switch {|| block} => object
*
* Switch the effective and real group IDs of the current process. If
* a <em>block</em> is given, the group IDs will be switched back
* after the block is executed. Returns the new effective group ID if
* called without a block, and the return value of the block if one
* is given.
*
*/
static VALUE
p_gid_switch(obj)
VALUE obj;
{
extern int errno;
int gid, egid;
check_gid_switch();
gid = getgid();
egid = getegid();
if (gid != egid) {
proc_setegid(obj, INT2FIX(gid));
if (rb_block_given_p()) {
under_gid_switch = 1;
return rb_ensure(rb_yield, Qnil, p_gid_sw_ensure, SAVED_GROUP_ID);
} else {
return INT2FIX(egid);
}
} else if (egid != SAVED_GROUP_ID) {
proc_setegid(obj, INT2FIX(SAVED_GROUP_ID));
if (rb_block_given_p()) {
under_gid_switch = 1;
return rb_ensure(rb_yield, Qnil, p_gid_sw_ensure, egid);
} else {
return INT2FIX(gid);
}
} else {
errno = EPERM;
rb_sys_fail(0);
}
#else
static VALUE
p_gid_sw_ensure(obj)
VALUE obj;
{
under_gid_switch = 0;
return p_gid_exchange(obj);
}
static VALUE
p_gid_switch(obj)
VALUE obj;
{
extern int errno;
int gid, egid;
check_gid_switch();
gid = getgid();
egid = getegid();
if (gid == egid) {
errno = EPERM;
rb_sys_fail(0);
}
p_gid_exchange(obj);
if (rb_block_given_p()) {
under_gid_switch = 1;
return rb_ensure(rb_yield, Qnil, p_gid_sw_ensure, obj);
} else {
return INT2FIX(egid);
}
#endif
}
/*
* call-seq:
* Process.times => aStructTms
*
* Returns a <code>Tms</code> structure (see <code>Struct::Tms</code>
* on page 388) that contains user and system CPU times for this
* process.
*
* t = Process.times
* [ t.utime, t.stime ] #=> [0.0, 0.02]
*/
VALUE
rb_proc_times(obj)
VALUE obj;
{
#if defined(HAVE_TIMES) && !defined(__CHECKER__)
#ifndef HZ
# ifdef CLK_TCK
# define HZ CLK_TCK
# else
# define HZ 60
# endif
#endif /* HZ */
struct tms buf;
volatile VALUE utime, stime, cutime, sctime;
times(&buf);
return rb_struct_new(S_Tms,
utime = rb_float_new((double)buf.tms_utime / HZ),
stime = rb_float_new((double)buf.tms_stime / HZ),
cutime = rb_float_new((double)buf.tms_cutime / HZ),
sctime = rb_float_new((double)buf.tms_cstime / HZ));
#else
rb_notimplement();
#endif
}
VALUE rb_mProcess;
VALUE rb_mProcUID;
VALUE rb_mProcGID;
VALUE rb_mProcID_Syscall;
/*
* The <code>Process</code> module is a collection of methods used to
* manipulate processes.
*/
void
Init_process()
{
rb_define_virtual_variable("$$", get_pid, 0);
rb_define_readonly_variable("$?", &rb_last_status);
rb_define_global_function("exec", rb_f_exec, -1);
rb_define_global_function("fork", rb_f_fork, 0);
rb_define_global_function("exit!", rb_f_exit_bang, -1);
rb_define_global_function("system", rb_f_system, -1);
rb_define_global_function("spawn", rb_f_spawn, -1);
rb_define_global_function("sleep", rb_f_sleep, -1);
rb_mProcess = rb_define_module("Process");
#if !defined(_WIN32) && !defined(DJGPP)
#ifdef WNOHANG
rb_define_const(rb_mProcess, "WNOHANG", INT2FIX(WNOHANG));
#else
rb_define_const(rb_mProcess, "WNOHANG", INT2FIX(0));
#endif
#ifdef WUNTRACED
rb_define_const(rb_mProcess, "WUNTRACED", INT2FIX(WUNTRACED));
#else
rb_define_const(rb_mProcess, "WUNTRACED", INT2FIX(0));
#endif
#endif
rb_define_singleton_method(rb_mProcess, "fork", rb_f_fork, 0);
rb_define_singleton_method(rb_mProcess, "spawn", rb_f_spawn, -1);
rb_define_singleton_method(rb_mProcess, "exit!", rb_f_exit_bang, -1);
rb_define_singleton_method(rb_mProcess, "exit", rb_f_exit, -1); /* in eval.c */
rb_define_singleton_method(rb_mProcess, "abort", rb_f_abort, -1); /* in eval.c */
rb_define_module_function(rb_mProcess, "kill", rb_f_kill, -1); /* in signal.c */
rb_define_module_function(rb_mProcess, "wait", proc_wait, -1);
rb_define_module_function(rb_mProcess, "wait2", proc_wait2, -1);
rb_define_module_function(rb_mProcess, "waitpid", proc_wait, -1);
rb_define_module_function(rb_mProcess, "waitpid2", proc_wait2, -1);
rb_define_module_function(rb_mProcess, "waitall", proc_waitall, 0);
rb_define_module_function(rb_mProcess, "detach", proc_detach, 1);
rb_cProcStatus = rb_define_class_under(rb_mProcess, "Status", rb_cObject);
rb_undef_method(CLASS_OF(rb_cProcStatus), "new");
rb_define_method(rb_cProcStatus, "==", pst_equal, 1);
rb_define_method(rb_cProcStatus, "&", pst_bitand, 1);
rb_define_method(rb_cProcStatus, ">>", pst_rshift, 1);
rb_define_method(rb_cProcStatus, "to_i", pst_to_i, 0);
rb_define_method(rb_cProcStatus, "to_int", pst_to_i, 0);
rb_define_method(rb_cProcStatus, "to_s", pst_to_s, 0);
rb_define_method(rb_cProcStatus, "inspect", pst_inspect, 0);
rb_define_method(rb_cProcStatus, "pid", pst_pid, 0);
rb_define_method(rb_cProcStatus, "stopped?", pst_wifstopped, 0);
rb_define_method(rb_cProcStatus, "stopsig", pst_wstopsig, 0);
rb_define_method(rb_cProcStatus, "signaled?", pst_wifsignaled, 0);
rb_define_method(rb_cProcStatus, "termsig", pst_wtermsig, 0);
rb_define_method(rb_cProcStatus, "exited?", pst_wifexited, 0);
rb_define_method(rb_cProcStatus, "exitstatus", pst_wexitstatus, 0);
rb_define_method(rb_cProcStatus, "success?", pst_success_p, 0);
rb_define_method(rb_cProcStatus, "coredump?", pst_wcoredump, 0);
rb_define_module_function(rb_mProcess, "pid", get_pid, 0);
rb_define_module_function(rb_mProcess, "ppid", get_ppid, 0);
rb_define_module_function(rb_mProcess, "getpgrp", proc_getpgrp, 0);
rb_define_module_function(rb_mProcess, "setpgrp", proc_setpgrp, 0);
rb_define_module_function(rb_mProcess, "getpgid", proc_getpgid, 1);
rb_define_module_function(rb_mProcess, "setpgid", proc_setpgid, 2);
rb_define_module_function(rb_mProcess, "setsid", proc_setsid, 0);
rb_define_module_function(rb_mProcess, "getpriority", proc_getpriority, 2);
rb_define_module_function(rb_mProcess, "setpriority", proc_setpriority, 3);
#ifdef HAVE_GETPRIORITY
rb_define_const(rb_mProcess, "PRIO_PROCESS", INT2FIX(PRIO_PROCESS));
rb_define_const(rb_mProcess, "PRIO_PGRP", INT2FIX(PRIO_PGRP));
rb_define_const(rb_mProcess, "PRIO_USER", INT2FIX(PRIO_USER));
#endif
#ifdef HAVE_GETRLIMIT
rb_define_module_function(rb_mProcess, "getrlimit", proc_getrlimit, 1);
#endif
#ifdef HAVE_SETRLIMIT
rb_define_module_function(rb_mProcess, "setrlimit", proc_setrlimit, 3);
#endif
#ifdef RLIM2NUM
#ifdef RLIM_INFINITY
rb_define_const(rb_mProcess, "RLIM_INFINITY", RLIM2NUM(RLIM_INFINITY));
#endif
#ifdef RLIM_SAVED_MAX
rb_define_const(rb_mProcess, "RLIM_SAVED_MAX", RLIM2NUM(RLIM_SAVED_MAX));
#endif
#ifdef RLIM_SAVED_CUR
rb_define_const(rb_mProcess, "RLIM_SAVED_CUR", RLIM2NUM(RLIM_SAVED_CUR));
#endif
#ifdef RLIMIT_CORE
rb_define_const(rb_mProcess, "RLIMIT_CORE", INT2FIX(RLIMIT_CORE));
#endif
#ifdef RLIMIT_CPU
rb_define_const(rb_mProcess, "RLIMIT_CPU", INT2FIX(RLIMIT_CPU));
#endif
#ifdef RLIMIT_DATA
rb_define_const(rb_mProcess, "RLIMIT_DATA", INT2FIX(RLIMIT_DATA));
#endif
#ifdef RLIMIT_FSIZE
rb_define_const(rb_mProcess, "RLIMIT_FSIZE", INT2FIX(RLIMIT_FSIZE));
#endif
#ifdef RLIMIT_NOFILE
rb_define_const(rb_mProcess, "RLIMIT_NOFILE", INT2FIX(RLIMIT_NOFILE));
#endif
#ifdef RLIMIT_STACK
rb_define_const(rb_mProcess, "RLIMIT_STACK", INT2FIX(RLIMIT_STACK));
#endif
#ifdef RLIMIT_AS
rb_define_const(rb_mProcess, "RLIMIT_AS", INT2FIX(RLIMIT_AS));
#endif
#ifdef RLIMIT_MEMLOCK
rb_define_const(rb_mProcess, "RLIMIT_MEMLOCK", INT2FIX(RLIMIT_MEMLOCK));
#endif
#ifdef RLIMIT_NPROC
rb_define_const(rb_mProcess, "RLIMIT_NPROC", INT2FIX(RLIMIT_NPROC));
#endif
#ifdef RLIMIT_RSS
rb_define_const(rb_mProcess, "RLIMIT_RSS", INT2FIX(RLIMIT_RSS));
#endif
#ifdef RLIMIT_SBSIZE
rb_define_const(rb_mProcess, "RLIMIT_SBSIZE", INT2FIX(RLIMIT_SBSIZE));
#endif
#endif
rb_define_module_function(rb_mProcess, "uid", proc_getuid, 0);
rb_define_module_function(rb_mProcess, "uid=", proc_setuid, 1);
rb_define_module_function(rb_mProcess, "gid", proc_getgid, 0);
rb_define_module_function(rb_mProcess, "gid=", proc_setgid, 1);
rb_define_module_function(rb_mProcess, "euid", proc_geteuid, 0);
rb_define_module_function(rb_mProcess, "euid=", proc_seteuid, 1);
rb_define_module_function(rb_mProcess, "egid", proc_getegid, 0);
rb_define_module_function(rb_mProcess, "egid=", proc_setegid, 1);
rb_define_module_function(rb_mProcess, "initgroups", proc_initgroups, 2);
rb_define_module_function(rb_mProcess, "groups", proc_getgroups, 0);
rb_define_module_function(rb_mProcess, "groups=", proc_setgroups, 1);
rb_define_module_function(rb_mProcess, "maxgroups", proc_getmaxgroups, 0);
rb_define_module_function(rb_mProcess, "maxgroups=", proc_setmaxgroups, 1);
rb_define_module_function(rb_mProcess, "daemon", proc_daemon, -1);
rb_define_module_function(rb_mProcess, "times", rb_proc_times, 0);
#if defined(HAVE_TIMES) || defined(_WIN32)
S_Tms = rb_struct_define("Tms", "utime", "stime", "cutime", "cstime", NULL);
#endif
SAVED_USER_ID = geteuid();
SAVED_GROUP_ID = getegid();
rb_mProcUID = rb_define_module_under(rb_mProcess, "UID");
rb_mProcGID = rb_define_module_under(rb_mProcess, "GID");
rb_define_module_function(rb_mProcUID, "rid", proc_getuid, 0);
rb_define_module_function(rb_mProcGID, "rid", proc_getgid, 0);
rb_define_module_function(rb_mProcUID, "eid", proc_geteuid, 0);
rb_define_module_function(rb_mProcGID, "eid", proc_getegid, 0);
rb_define_module_function(rb_mProcUID, "change_privilege", p_uid_change_privilege, 1);
rb_define_module_function(rb_mProcGID, "change_privilege", p_gid_change_privilege, 1);
rb_define_module_function(rb_mProcUID, "grant_privilege", p_uid_grant_privilege, 1);
rb_define_module_function(rb_mProcGID, "grant_privilege", p_gid_grant_privilege, 1);
rb_define_alias(rb_mProcUID, "eid=", "grant_privilege");
rb_define_alias(rb_mProcGID, "eid=", "grant_privilege");
rb_define_module_function(rb_mProcUID, "re_exchange", p_uid_exchange, 0);
rb_define_module_function(rb_mProcGID, "re_exchange", p_gid_exchange, 0);
rb_define_module_function(rb_mProcUID, "re_exchangeable?", p_uid_exchangeable, 0);
rb_define_module_function(rb_mProcGID, "re_exchangeable?", p_gid_exchangeable, 0);
rb_define_module_function(rb_mProcUID, "sid_available?", p_uid_have_saved_id, 0);
rb_define_module_function(rb_mProcGID, "sid_available?", p_gid_have_saved_id, 0);
rb_define_module_function(rb_mProcUID, "switch", p_uid_switch, 0);
rb_define_module_function(rb_mProcGID, "switch", p_gid_switch, 0);
rb_mProcID_Syscall = rb_define_module_under(rb_mProcess, "Sys");
rb_define_module_function(rb_mProcID_Syscall, "getuid", proc_getuid, 0);
rb_define_module_function(rb_mProcID_Syscall, "geteuid", proc_geteuid, 0);
rb_define_module_function(rb_mProcID_Syscall, "getgid", proc_getgid, 0);
rb_define_module_function(rb_mProcID_Syscall, "getegid", proc_getegid, 0);
rb_define_module_function(rb_mProcID_Syscall, "setuid", p_sys_setuid, 1);
rb_define_module_function(rb_mProcID_Syscall, "setgid", p_sys_setgid, 1);
rb_define_module_function(rb_mProcID_Syscall, "setruid", p_sys_setruid, 1);
rb_define_module_function(rb_mProcID_Syscall, "setrgid", p_sys_setrgid, 1);
rb_define_module_function(rb_mProcID_Syscall, "seteuid", p_sys_seteuid, 1);
rb_define_module_function(rb_mProcID_Syscall, "setegid", p_sys_setegid, 1);
rb_define_module_function(rb_mProcID_Syscall, "setreuid", p_sys_setreuid, 2);
rb_define_module_function(rb_mProcID_Syscall, "setregid", p_sys_setregid, 2);
rb_define_module_function(rb_mProcID_Syscall, "setresuid", p_sys_setresuid, 3);
rb_define_module_function(rb_mProcID_Syscall, "setresgid", p_sys_setresgid, 3);
rb_define_module_function(rb_mProcID_Syscall, "issetugid", p_sys_issetugid, 0);
}