[dragonfly.git] / contrib / tcsh-6 / sh.proc.c

/* $Header: /p/tcsh/cvsroot/tcsh/sh.proc.c,v 3.104 2006/09/27 16:59:04 mitr Exp $ */
/*
 * sh.proc.c: Job manipulations
 */
/*-
 * Copyright (c) 1980, 1991 The Regents of the University of California.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */
#include "sh.h"

RCSID("$tcsh: sh.proc.c,v 3.104 2006/09/27 16:59:04 mitr Exp $")

#include "ed.h"
#include "tc.h"
#include "tc.wait.h"

#ifdef WINNT_NATIVE
#undef POSIX
#define POSIX
#endif /* WINNT_NATIVE */
#ifdef aiws
# undef HZ
# define HZ 16
#endif /* aiws */

#if defined(_BSD) || (defined(IRIS4D) && __STDC__) || defined(__lucid) || defined(linux) || defined(__GNU__) || defined(__GLIBC__)
# define BSDWAIT
#endif /* _BSD || (IRIS4D && __STDC__) || __lucid || glibc */
#ifndef WTERMSIG
# define WTERMSIG(w)    (((union wait *) &(w))->w_termsig)
# ifndef BSDWAIT
#  define BSDWAIT
# endif /* !BSDWAIT */
#endif /* !WTERMSIG */
#ifndef WEXITSTATUS
# define WEXITSTATUS(w) (((union wait *) &(w))->w_retcode)
#endif /* !WEXITSTATUS */
#ifndef WSTOPSIG
# define WSTOPSIG(w)    (((union wait *) &(w))->w_stopsig)
#endif /* !WSTOPSIG */

#ifdef __osf__
# ifndef WCOREDUMP
#  define WCOREDUMP(x) (_W_INT(x) & WCOREFLAG)
# endif
#endif

#ifndef WCOREDUMP
# ifdef BSDWAIT
#  define WCOREDUMP(w)  (((union wait *) &(w))->w_coredump)
# else /* !BSDWAIT */
#  define WCOREDUMP(w)  ((w) & 0200)
# endif /* !BSDWAIT */
#endif /* !WCOREDUMP */

#ifndef JOBDEBUG
# define jobdebug_xprintf(x)    (void)0
# define jobdebug_flush()       (void)0
#else
# define jobdebug_xprintf(s)    xprintf s
# define jobdebug_flush()       flush()
#endif

/*
 * C Shell - functions that manage processes, handling hanging, termination
 */

#define BIGINDEX        9       /* largest desirable job index */

#ifdef BSDTIMES
# ifdef convex
/* use 'cvxrusage' to get parallel statistics */
static struct cvxrusage zru = {{0L, 0L}, {0L, 0L}, 0L, 0L, 0L, 0L,
                                0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
                                {0L, 0L}, 0LL, 0LL, 0LL, 0LL, 0L, 0L, 0L,
                                0LL, 0LL, {0L, 0L, 0L, 0L, 0L}};
# else
static struct rusage zru;
# endif /* convex */
#else /* !BSDTIMES */
# ifdef _SEQUENT_
static struct process_stats zru = {{0L, 0L}, {0L, 0L}, 0, 0, 0, 0, 0, 0, 0,
                                   0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
# else /* !_SEQUENT_ */
#  ifdef _SX
static struct tms zru = {0, 0, 0, 0}, lru = {0, 0, 0, 0};
#  else /* !_SX */
static struct tms zru = {0L, 0L, 0L, 0L}, lru = {0L, 0L, 0L, 0L};
#  endif        /* !_SX */
# endif /* !_SEQUENT_ */
#endif /* !BSDTIMES */

#ifndef BSDTIMES
static int timesdone;   /* shtimes buffer full ? */
#endif /* BSDTIMES */

#ifndef RUSAGE_CHILDREN
# define        RUSAGE_CHILDREN -1
#endif /* RUSAGE_CHILDREN */

static  void             pflushall      (void);
static  void             pflush         (struct process *);
static  void             pfree          (struct process *);
static  void             pclrcurr       (struct process *);
static  void             padd           (struct command *);
static  int              pprint         (struct process *, int);
static  void             ptprint        (struct process *);
static  void             pads           (Char *);
static  void             pkill          (Char **, int);
static  struct process  *pgetcurr       (struct process *);
static  void             okpcntl        (void);
static  void             setttypgrp     (int);

/*
 * pchild - call queued by the SIGCHLD signal
 *      indicating that at least one child has terminated or stopped
 *      thus at least one wait system call will definitely return a
 *      childs status.  Top level routines (like pwait) must be sure
 *      to mask interrupts when playing with the proclist data structures!
 */
void
pchild(void)
{
    struct process *pp;
    struct process *fp;
    pid_t pid;
#ifdef BSDWAIT
    union wait w;
#else /* !BSDWAIT */
    int     w;
#endif /* !BSDWAIT */
    int     jobflags;
#ifdef BSDTIMES
    struct sysrusage ru;
#else /* !BSDTIMES */
# ifdef _SEQUENT_
    struct process_stats ru;
    struct process_stats cpst1, cpst2;
    timeval_t tv;
# else /* !_SEQUENT_ */
    struct tms proctimes;

    if (!timesdone) {
        timesdone++;
        (void) times(&shtimes);
    }
# endif /* !_SEQUENT_ */
#endif /* !BSDTIMES */

    jobdebug_xprintf(("pchild()\n"));

loop:
    jobdebug_xprintf(("Waiting...\n"));
    jobdebug_flush();
    errno = 0;                  /* reset, just in case */

#ifndef WINNT_NATIVE
# ifdef BSDJOBS
#  ifdef BSDTIMES
#   ifdef convex
    /* use 'cvxwait' to get parallel statistics */
    pid = cvxwait(&w,
        (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG), &ru);
#   else
    /* both a wait3 and rusage */
#    if !defined(BSDWAIT) || defined(NeXT) || defined(MACH) || defined(linux) || defined(__GNU__) || defined(__GLIBC__) || (defined(IRIS4D) && SYSVREL <= 3) || defined(__lucid) || defined(__osf__)
    pid = wait3(&w,
       (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG), &ru);
#    else /* BSDWAIT */
    pid = wait3(&w.w_status,
       (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG), &ru);
#    endif /* BSDWAIT */
#   endif /* convex */
#  else /* !BSDTIMES */
#   ifdef _SEQUENT_
    (void) get_process_stats(&tv, PS_SELF, 0, &cpst1);
    pid = waitpid(-1, &w,
            (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG));
    (void) get_process_stats(&tv, PS_SELF, 0, &cpst2);
    pr_stat_sub(&cpst2, &cpst1, &ru);
#   else        /* !_SEQUENT_ */
#    ifndef POSIX
    /* we have a wait3, but no rusage stuff */
    pid = wait3(&w.w_status,
         (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG), 0);
#    else /* POSIX */
    pid = waitpid(-1, &w,
            (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG));
#    endif /* POSIX */
#   endif /* !_SEQUENT_ */
#  endif        /* !BSDTIMES */
# else /* !BSDJOBS */
#  ifdef BSDTIMES
#   define HAVEwait3
    /* both a wait3 and rusage */
#   ifdef hpux
    pid = wait3(&w.w_status, WNOHANG, 0);
#   else        /* !hpux */
    pid = wait3(&w.w_status, WNOHANG, &ru);
#   endif /* !hpux */
#  else /* !BSDTIMES */
#   ifdef ODT  /* For Sco Unix 3.2.0 or ODT 1.0 */
#    define HAVEwait3
    pid = waitpid(-1, &w,
            (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG));
#   endif /* ODT */         
#   if defined(aiws) || defined(uts)
#    define HAVEwait3
    pid = wait3(&w.w_status, 
        (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG), 0);
#   endif /* aiws || uts */
#   ifndef HAVEwait3
#    ifndef BSDWAIT
     /* no wait3, therefore no rusage */
     /* on Sys V, this may hang.  I hope it's not going to be a problem */
    pid = wait(&w);
#    else /* BSDWAIT */
     /* 
      * XXX: for greater than 3 we should use waitpid(). 
      * but then again, SVR4 falls into the POSIX/BSDJOBS category.
      */
    pid = wait(&w.w_status);
#    endif /* BSDWAIT */
#   endif /* !HAVEwait3 */
#  endif        /* !BSDTIMES */
# endif /* !BSDJOBS */
#else /* WINNT_NATIVE */
    pid = waitpid(-1, &w,
            (setintr && (intty || insource) ? WNOHANG | WUNTRACED : WNOHANG));
#endif /* WINNT_NATIVE */

    jobdebug_xprintf(("parent %d pid %d, retval %x termsig %x retcode %x\n",
                      (int)getpid(), (int)pid, w, WTERMSIG(w),
                      WEXITSTATUS(w)));
    jobdebug_flush();

    if ((pid == 0) || (pid == -1)) {
        handle_pending_signals();
        jobdebug_xprintf(("errno == %d\n", errno));
        if (errno == EINTR)
            goto loop;
        goto end;
    }
    for (pp = proclist.p_next; pp != NULL; pp = pp->p_next)
        if (pid == pp->p_procid)
            goto found;
#if !defined(BSDJOBS) && !defined(WINNT_NATIVE)
    /* this should never have happened */
    stderror(ERR_SYNC, pid);
    xexit(0);
#else /* BSDJOBS || WINNT_NATIVE */
    goto loop;
#endif /* !BSDJOBS && !WINNT_NATIVE */
found:
    pp->p_flags &= ~(PRUNNING | PSTOPPED | PREPORTED);
    if (WIFSTOPPED(w)) {
        pp->p_flags |= PSTOPPED;
        pp->p_reason = WSTOPSIG(w);
    }
    else {
        if (pp->p_flags & (PTIME | PPTIME) || adrof(STRtime))
#ifndef BSDTIMES
# ifdef _SEQUENT_
            (void) get_process_stats(&pp->p_etime, PS_SELF, NULL, NULL);
# else  /* !_SEQUENT_ */
            pp->p_etime = times(&proctimes);
# endif /* !_SEQUENT_ */
#else /* BSDTIMES */
            (void) gettimeofday(&pp->p_etime, NULL);
#endif /* BSDTIMES */


#if defined(BSDTIMES) || defined(_SEQUENT_)
        pp->p_rusage = ru;
#else /* !BSDTIMES && !_SEQUENT_ */
        (void) times(&proctimes);
        pp->p_utime = proctimes.tms_cutime - shtimes.tms_cutime;
        pp->p_stime = proctimes.tms_cstime - shtimes.tms_cstime;
        shtimes = proctimes;
#endif /* !BSDTIMES && !_SEQUENT_ */
        if (WIFSIGNALED(w)) {
            if (WTERMSIG(w) == SIGINT)
                pp->p_flags |= PINTERRUPTED;
            else
                pp->p_flags |= PSIGNALED;
            if (WCOREDUMP(w))
                pp->p_flags |= PDUMPED;
            pp->p_reason = WTERMSIG(w);
        }
        else {
            pp->p_reason = WEXITSTATUS(w);
            if (pp->p_reason != 0)
                pp->p_flags |= PAEXITED;
            else
                pp->p_flags |= PNEXITED;
        }
    }
    jobflags = 0;
    fp = pp;
    do {
        if ((fp->p_flags & (PPTIME | PRUNNING | PSTOPPED)) == 0 &&
            !child && adrof(STRtime) &&
#ifdef BSDTIMES
            fp->p_rusage.ru_utime.tv_sec + fp->p_rusage.ru_stime.tv_sec
#else /* !BSDTIMES */
# ifdef _SEQUENT_
            fp->p_rusage.ps_utime.tv_sec + fp->p_rusage.ps_stime.tv_sec
# else /* !_SEQUENT_ */
#  ifndef POSIX
            (fp->p_utime + fp->p_stime) / HZ
#  else /* POSIX */
            (fp->p_utime + fp->p_stime) / clk_tck
#  endif /* POSIX */
# endif /* !_SEQUENT_ */
#endif /* !BSDTIMES */
            >= atoi(short2str(varval(STRtime))))
            fp->p_flags |= PTIME;
        jobflags |= fp->p_flags;
    } while ((fp = fp->p_friends) != pp);
    pp->p_flags &= ~PFOREGND;
    if (pp == pp->p_friends && (pp->p_flags & PPTIME)) {
        pp->p_flags &= ~PPTIME;
        pp->p_flags |= PTIME;
    }
    if ((jobflags & (PRUNNING | PREPORTED)) == 0) {
        fp = pp;
        do {
            if (fp->p_flags & PSTOPPED)
                fp->p_flags |= PREPORTED;
        } while ((fp = fp->p_friends) != pp);
        while (fp->p_procid != fp->p_jobid)
            fp = fp->p_friends;
        if (jobflags & PSTOPPED) {
            if (pcurrent && pcurrent != fp)
                pprevious = pcurrent;
            pcurrent = fp;
        }
        else
            pclrcurr(fp);
        if (jobflags & PFOREGND) {
            if (!(jobflags & (PSIGNALED | PSTOPPED | PPTIME) ||
#ifdef notdef
                jobflags & PAEXITED ||
#endif /* notdef */
                !eq(dcwd->di_name, fp->p_cwd->di_name))) {
            /* PWP: print a newline after ^C */
                if (jobflags & PINTERRUPTED) {
                    xputchar('\r' | QUOTE);
                    xputchar('\n');
                }
#ifdef notdef
                else if ((jobflags & (PTIME|PSTOPPED)) == PTIME)
                    ptprint(fp);
#endif /* notdef */
            }
        }
        else {
            if (jobflags & PNOTIFY || adrof(STRnotify)) {
                xputchar('\r' | QUOTE);
                xputchar('\n');
                (void) pprint(pp, NUMBER | NAME | REASON);
                if ((jobflags & PSTOPPED) == 0)
                    pflush(pp);
                if (GettingInput) {
                    errno = 0;
                    (void) Rawmode();
#ifdef notdef
                    /*
                     * don't really want to do that, because it
                     * will erase our message in case of multi-line
                     * input
                     */
                    ClearLines();
#endif /* notdef */
                    ClearDisp();
                    Refresh();
                }
            }
            else {
                fp->p_flags |= PNEEDNOTE;
                neednote = 1;
            }
        }
    }
#if defined(BSDJOBS) || defined(HAVEwait3) ||defined(WINNT_NATIVE)
    goto loop;
#endif /* BSDJOBS || HAVEwait3 */
 end:
    ;
}

void
pnote(void)
{
    struct process *pp;
    int     flags;

    neednote = 0;
    for (pp = proclist.p_next; pp != NULL; pp = pp->p_next) {
        if (pp->p_flags & PNEEDNOTE) {
            pchild_disabled++;
            cleanup_push(&pchild_disabled, disabled_cleanup);
            pp->p_flags &= ~PNEEDNOTE;
            flags = pprint(pp, NUMBER | NAME | REASON);
            if ((flags & (PRUNNING | PSTOPPED)) == 0)
                pflush(pp);
            cleanup_until(&pchild_disabled);
        }
    }
}


static void
pfree(struct process *pp)
{       
    xfree(pp->p_command);
    if (pp->p_cwd && --pp->p_cwd->di_count == 0)
        if (pp->p_cwd->di_next == 0)
            dfree(pp->p_cwd);
    xfree(pp);
}


/*
 * pwait - wait for current job to terminate, maintaining integrity
 *      of current and previous job indicators.
 */
void
pwait(void)
{
    struct process *fp, *pp;

    /*
     * Here's where dead procs get flushed.
     */
    for (pp = (fp = &proclist)->p_next; pp != NULL; pp = (fp = pp)->p_next)
        if (pp->p_procid == 0) {
            fp->p_next = pp->p_next;
            pfree(pp);
            pp = fp;
        }
    pjwait(pcurrjob);
}


/*
 * pjwait - wait for a job to finish or become stopped
 *      It is assumed to be in the foreground state (PFOREGND)
 */
void
pjwait(struct process *pp)
{
    struct process *fp;
    int     jobflags, reason;
    sigset_t oset, set, pause_mask;
    Char *reason_str;

    while (pp->p_procid != pp->p_jobid)
        pp = pp->p_friends;
    fp = pp;

    do {
        if ((fp->p_flags & (PFOREGND | PRUNNING)) == PRUNNING)
          xprintf(CGETS(17, 1, "BUG: waiting for background job!\n"));
    } while ((fp = fp->p_friends) != pp);
    /*
     * Now keep pausing as long as we are not interrupted (SIGINT), and the
     * target process, or any of its friends, are running
     */
    fp = pp;
    sigemptyset(&set);
    sigaddset(&set, SIGINT);
    sigaddset(&set, SIGCHLD);
    (void)sigprocmask(SIG_BLOCK, &set, &oset);
    cleanup_push(&oset, sigprocmask_cleanup);
    pause_mask = oset;
    sigdelset(&pause_mask, SIGCHLD);
    for (;;) {
        handle_pending_signals();
        jobflags = 0;
        do
            jobflags |= fp->p_flags;
        while ((fp = (fp->p_friends)) != pp);
        if ((jobflags & PRUNNING) == 0)
            break;
        jobdebug_xprintf(("%d starting to sigsuspend for SIGCHLD on %d\n",
                          getpid(), fp->p_procid));
        sigsuspend(&pause_mask);
    }
    cleanup_until(&oset);
    jobdebug_xprintf(("%d returned from sigsuspend loop\n", getpid()));
#ifdef BSDJOBS
    if (tpgrp > 0)              /* get tty back */
        (void) tcsetpgrp(FSHTTY, tpgrp);
#endif /* BSDJOBS */
    if ((jobflags & (PSIGNALED | PSTOPPED | PTIME)) ||
        !eq(dcwd->di_name, fp->p_cwd->di_name)) {
        if (jobflags & PSTOPPED) {
            xputchar('\n');
            if (adrof(STRlistjobs)) {
                Char   *jobcommand[3];

                jobcommand[0] = STRjobs;
                if (eq(varval(STRlistjobs), STRlong))
                    jobcommand[1] = STRml;
                else
                    jobcommand[1] = NULL;
                jobcommand[2] = NULL;

                dojobs(jobcommand, NULL);
                (void) pprint(pp, SHELLDIR);
            }
            else
                (void) pprint(pp, AREASON | SHELLDIR);
        }
        else
            (void) pprint(pp, AREASON | SHELLDIR);
    }
    if ((jobflags & (PINTERRUPTED | PSTOPPED)) && setintr &&
        (!gointr || !eq(gointr, STRminus))) {
        if ((jobflags & PSTOPPED) == 0)
            pflush(pp);
        pintr1(0);
        /* NOTREACHED */
    }
    reason = 0;
    fp = pp;
    do {
        if (fp->p_reason)
            reason = fp->p_flags & (PSIGNALED | PINTERRUPTED) ?
                fp->p_reason | META : fp->p_reason;
    } while ((fp = fp->p_friends) != pp);
    /*
     * Don't report on backquoted jobs, cause it will mess up 
     * their output.
     */
    if ((reason != 0) && (adrof(STRprintexitvalue)) && 
        (pp->p_flags & PBACKQ) == 0)
        xprintf(CGETS(17, 2, "Exit %d\n"), reason);
    reason_str = putn(reason);
    cleanup_push(reason_str, xfree);
    setv(STRstatus, reason_str, VAR_READWRITE);
    cleanup_ignore(reason_str);
    cleanup_until(reason_str);
    if (reason && exiterr)
        exitstat();
    pflush(pp);
}

/*
 * dowait - wait for all processes to finish
 */

/*ARGSUSED*/
void
dowait(Char **v, struct command *c)
{
    struct process *pp;
    sigset_t pause_mask;

    USE(c);
    USE(v);
    pjobs++;
    sigprocmask(SIG_BLOCK, NULL, &pause_mask);
    sigdelset(&pause_mask, SIGCHLD);
    if (setintr)
        sigdelset(&pause_mask, SIGINT);
loop:
    for (pp = proclist.p_next; pp; pp = pp->p_next)
        if (pp->p_procid &&     /* pp->p_procid == pp->p_jobid && */
            pp->p_flags & PRUNNING) {
            handle_pending_signals();
            sigsuspend(&pause_mask);
            handle_pending_signals();
            goto loop;
        }
    pjobs = 0;
}

/*
 * pflushall - flush all jobs from list (e.g. at fork())
 */
static void
pflushall(void)
{
    struct process *pp;

    for (pp = proclist.p_next; pp != NULL; pp = pp->p_next)
        if (pp->p_procid)
            pflush(pp);
}

/*
 * pflush - flag all process structures in the same job as the
 *      the argument process for deletion.  The actual free of the
 *      space is not done here since pflush is called at interrupt level.
 */
static void
pflush(struct process *pp)
{
    struct process *np;
    int idx;

    if (pp->p_procid == 0) {
        xprintf(CGETS(17, 3, "BUG: process flushed twice"));
        return;
    }
    while (pp->p_procid != pp->p_jobid)
        pp = pp->p_friends;
    pclrcurr(pp);
    if (pp == pcurrjob)
        pcurrjob = 0;
    idx = pp->p_index;
    np = pp;
    do {
        np->p_index = np->p_procid = 0;
        np->p_flags &= ~PNEEDNOTE;
    } while ((np = np->p_friends) != pp);
    if (idx == pmaxindex) {
        for (np = proclist.p_next, idx = 0; np; np = np->p_next)
            if (np->p_index > idx)
                idx = np->p_index;
        pmaxindex = idx;
    }
}

/*
 * pclrcurr - make sure the given job is not the current or previous job;
 *      pp MUST be the job leader
 */
static void
pclrcurr(struct process *pp)
{
    if (pp == pcurrent) {
        if (pprevious != NULL) {
            pcurrent = pprevious;
            pprevious = pgetcurr(pp);
        }
        else {
            pcurrent = pgetcurr(pp);
            pprevious = pgetcurr(pp);
        }
    }
    else if (pp == pprevious)
        pprevious = pgetcurr(pp);
}

/* +4 here is 1 for '\0', 1 ea for << >& >> */
static Char command[PMAXLEN + 4];
static size_t cmdlen;
static Char *cmdp;

/* GrP
 * unparse - Export padd() functionality 
 */
Char *
unparse(struct command *t)
{
    cmdp = command;
    cmdlen = 0;
    padd(t);
    *cmdp++ = '\0';
    return Strsave(command);
}


/*
 * palloc - allocate a process structure and fill it up.
 *      an important assumption is made that the process is running.
 */
void
palloc(pid_t pid, struct command *t)
{
    struct process *pp;
    int     i;

    pp = xcalloc(1, sizeof(struct process));
    pp->p_procid = pid;
    pp->p_flags = ((t->t_dflg & F_AMPERSAND) ? 0 : PFOREGND) | PRUNNING;
    if (t->t_dflg & F_TIME)
        pp->p_flags |= PPTIME;
    if (t->t_dflg & F_BACKQ)
        pp->p_flags |= PBACKQ;
    if (t->t_dflg & F_HUP)
        pp->p_flags |= PHUP;
    cmdp = command;
    cmdlen = 0;
    padd(t);
    *cmdp++ = 0;
    if (t->t_dflg & F_PIPEOUT) {
        pp->p_flags |= PPOU;
        if (t->t_dflg & F_STDERR)
            pp->p_flags |= PDIAG;
    }
    pp->p_command = Strsave(command);
    if (pcurrjob) {
        struct process *fp;

        /* careful here with interrupt level */
        pp->p_cwd = 0;
        pp->p_index = pcurrjob->p_index;
        pp->p_friends = pcurrjob;
        pp->p_jobid = pcurrjob->p_procid;
        for (fp = pcurrjob; fp->p_friends != pcurrjob; fp = fp->p_friends)
            continue;
        fp->p_friends = pp;
    }
    else {
        pcurrjob = pp;
        pp->p_jobid = pid;
        pp->p_friends = pp;
        pp->p_cwd = dcwd;
        dcwd->di_count++;
        if (pmaxindex < BIGINDEX)
            pp->p_index = ++pmaxindex;
        else {
            struct process *np;

            for (i = 1;; i++) {
                for (np = proclist.p_next; np; np = np->p_next)
                    if (np->p_index == i)
                        goto tryagain;
                pp->p_index = i;
                if (i > pmaxindex)
                    pmaxindex = i;
                break;
        tryagain:;
            }
        }
        if (pcurrent == NULL)
            pcurrent = pp;
        else if (pprevious == NULL)
            pprevious = pp;
    }
    pp->p_next = proclist.p_next;
    proclist.p_next = pp;
#ifdef BSDTIMES
    (void) gettimeofday(&pp->p_btime, NULL);
#else /* !BSDTIMES */
# ifdef _SEQUENT_
    (void) get_process_stats(&pp->p_btime, PS_SELF, NULL, NULL);
# else /* !_SEQUENT_ */
    {
        struct tms tmptimes;

        pp->p_btime = times(&tmptimes);
    }
# endif /* !_SEQUENT_ */
#endif /* !BSDTIMES */
}

static void
padd(struct command *t)
{
    Char  **argp;

    if (t == 0)
        return;
    switch (t->t_dtyp) {

    case NODE_PAREN:
        pads(STRLparensp);
        padd(t->t_dspr);
        pads(STRspRparen);
        break;

    case NODE_COMMAND:
        for (argp = t->t_dcom; *argp; argp++) {
            pads(*argp);
            if (argp[1])
                pads(STRspace);
        }
        break;

    case NODE_OR:
    case NODE_AND:
    case NODE_PIPE:
    case NODE_LIST:
        padd(t->t_dcar);
        switch (t->t_dtyp) {
        case NODE_OR:
            pads(STRspor2sp);
            break;
        case NODE_AND:
            pads(STRspand2sp);
            break;
        case NODE_PIPE:
            pads(STRsporsp);
            break;
        case NODE_LIST:
            pads(STRsemisp);
            break;
        default:
            break;
        }
        padd(t->t_dcdr);
        return;

    default:
        break;
    }
    if ((t->t_dflg & F_PIPEIN) == 0 && t->t_dlef) {
        pads((t->t_dflg & F_READ) ? STRspLarrow2sp : STRspLarrowsp);
        pads(t->t_dlef);
    }
    if ((t->t_dflg & F_PIPEOUT) == 0 && t->t_drit) {
        pads((t->t_dflg & F_APPEND) ? STRspRarrow2 : STRspRarrow);
        if (t->t_dflg & F_STDERR)
            pads(STRand);
        pads(STRspace);
        pads(t->t_drit);
    }
}

static void
pads(Char *cp)
{
    size_t i;

    /*
     * Avoid the Quoted Space alias hack! Reported by:
     * sam@john-bigboote.ICS.UCI.EDU (Sam Horrocks)
     */
    if (cp[0] == STRQNULL[0])
        cp++;

    i = Strlen(cp);

    if (cmdlen >= PMAXLEN)
        return;
    if (cmdlen + i >= PMAXLEN) {
        (void) Strcpy(cmdp, STRsp3dots);
        cmdlen = PMAXLEN;
        cmdp += 4;
        return;
    }
    (void) Strcpy(cmdp, cp);
    cmdp += i;
    cmdlen += i;
}

/*
 * psavejob - temporarily save the current job on a one level stack
 *      so another job can be created.  Used for { } in exp6
 *      and `` in globbing.
 */
void
psavejob(void)
{
    pholdjob = pcurrjob;
    pcurrjob = NULL;
}

void
psavejob_cleanup(void *dummy)
{
    USE(dummy);
    pcurrjob = pholdjob;
    pholdjob = NULL;
}

/*
 * pendjob - indicate that a job (set of commands) has been completed
 *      or is about to begin.
 */
void
pendjob(void)
{
    struct process *pp, *tp;

    if (pcurrjob && (pcurrjob->p_flags & (PFOREGND | PSTOPPED)) == 0) {
        pp = pcurrjob;
        while (pp->p_procid != pp->p_jobid)
            pp = pp->p_friends;
        xprintf("[%d]", pp->p_index);
        tp = pp;
        do {
            xprintf(" %d", pp->p_procid);
            pp = pp->p_friends;
        } while (pp != tp);
        xputchar('\n');
    }
    pholdjob = pcurrjob = 0;
}

/*
 * pprint - print a job
 */

/*
 * Hacks have been added for SVR4 to deal with pipe's being spawned in
 * reverse order
 *
 * David Dawes (dawes@physics.su.oz.au) Oct 1991
 */

static int
pprint(struct process *pp, int flag)
{
    int status, reason;
    struct process *tp;
    int     jobflags, pstatus, pcond;
    const char *format;

#ifdef BACKPIPE
    struct process *pipehead = NULL, *pipetail = NULL, *pmarker = NULL;
    int inpipe = 0;
#endif /* BACKPIPE */

    while (pp->p_procid != pp->p_jobid)
        pp = pp->p_friends;
    if (pp == pp->p_friends && (pp->p_flags & PPTIME)) {
        pp->p_flags &= ~PPTIME;
        pp->p_flags |= PTIME;
    }
    tp = pp;
    status = reason = -1;
    jobflags = 0;
    do {
#ifdef BACKPIPE
        /*
         * The pipeline is reversed, so locate the real head of the pipeline
         * if pp is at the tail of a pipe (and not already in a pipeline)
         */
        if ((pp->p_friends->p_flags & PPOU) && !inpipe && (flag & NAME)) {
            inpipe = 1;
            pipetail = pp;
            do 
                pp = pp->p_friends;
            while (pp->p_friends->p_flags & PPOU);
            pipehead = pp;
            pmarker = pp;
        /*
         * pmarker is used to hold the place of the proc being processed, so
         * we can search for the next one downstream later.
         */
        }
        pcond = (tp != pp || (inpipe && tp == pp));
#else /* !BACKPIPE */
        pcond = (tp != pp);
#endif /* BACKPIPE */       

        jobflags |= pp->p_flags;
        pstatus = (int) (pp->p_flags & PALLSTATES);
        if (pcond && linp != linbuf && !(flag & FANCY) &&
            ((pstatus == status && pp->p_reason == reason) ||
             !(flag & REASON)))
            xputchar(' ');
        else {
            if (pcond && linp != linbuf)
                xputchar('\n');
            if (flag & NUMBER) {
#ifdef BACKPIPE
                pcond = ((pp == tp && !inpipe) ||
                         (inpipe && pipetail == tp && pp == pipehead));
#else /* BACKPIPE */
                pcond = (pp == tp);
#endif /* BACKPIPE */
                if (pcond)
                    xprintf("[%d]%s %c ", pp->p_index,
                            pp->p_index < 10 ? " " : "",
                            pp == pcurrent ? '+' :
                            (pp == pprevious ? '-' : ' '));
                else
                    xprintf("       ");
            }
            if (flag & FANCY) {
                xprintf("%5d ", pp->p_procid);
#ifdef TCF
                xprintf("%11s ", sitename(pp->p_procid));
#endif /* TCF */
            }
            if (flag & (REASON | AREASON)) {
                if (flag & NAME)
                    format = "%-30s";
                else
                    format = "%s";
                if (pstatus == status) {
                    if (pp->p_reason == reason) {
                        xprintf(format, "");
                        goto prcomd;
                    }
                    else
                        reason = (int) pp->p_reason;
                }
                else {
                    status = pstatus;
                    reason = (int) pp->p_reason;
                }
                switch (status) {

                case PRUNNING:
                    xprintf(format, CGETS(17, 4, "Running "));
                    break;

                case PINTERRUPTED:
                case PSTOPPED:
                case PSIGNALED:
                    /*
                     * tell what happened to the background job
                     * From: Michael Schroeder
                     * <mlschroe@immd4.informatik.uni-erlangen.de>
                     */
                    if ((flag & REASON)
                        || ((flag & AREASON)
                            && reason != SIGINT
                            && (reason != SIGPIPE
                                || (pp->p_flags & PPOU) == 0))) {
                        char *ptr;
                        int free_ptr;

                        free_ptr = 0;
                        ptr = (char *)(intptr_t)mesg[pp->p_reason & 0177].pname;
                        if (ptr == NULL) {
                            ptr = xasprintf("%s %d", CGETS(17, 5, "Signal"),
                                            pp->p_reason & 0177);
                            cleanup_push(ptr, xfree);
                            free_ptr = 1;
                        }
                        xprintf(format, ptr);
                        if (free_ptr != 0)
                            cleanup_until(ptr);
                    }
                    else
                        reason = -1;
                    break;

                case PNEXITED:
                case PAEXITED:
                    if (flag & REASON) {
                        if (pp->p_reason)
                            xprintf(CGETS(17, 6, "Exit %-25d"), pp->p_reason);
                        else
                            xprintf(format, CGETS(17, 7, "Done"));
                    }
                    break;

                default:
                    xprintf(CGETS(17, 8, "BUG: status=%-9o"),
                            status);
                }
            }
        }
prcomd:
        if (flag & NAME) {
            xprintf("%S", pp->p_command);
            if (pp->p_flags & PPOU)
                xprintf(" |");
            if (pp->p_flags & PDIAG)
                xprintf("&");
        }
        if (flag & (REASON | AREASON) && pp->p_flags & PDUMPED)
            xprintf(CGETS(17, 9, " (core dumped)"));
        if (tp == pp->p_friends) {
            if (flag & AMPERSAND)
                xprintf(" &");
            if (flag & JOBDIR &&
                !eq(tp->p_cwd->di_name, dcwd->di_name)) {
                xprintf(CGETS(17, 10, " (wd: "));
                dtildepr(tp->p_cwd->di_name);
                xprintf(")");
            }
        }
        if (pp->p_flags & PPTIME && !(status & (PSTOPPED | PRUNNING))) {
            if (linp != linbuf)
                xprintf("\n\t");
#if defined(BSDTIMES) || defined(_SEQUENT_)
            prusage(&zru, &pp->p_rusage, &pp->p_etime,
                    &pp->p_btime);
#else /* !BSDTIMES && !SEQUENT */
            lru.tms_utime = pp->p_utime;
            lru.tms_stime = pp->p_stime;
            lru.tms_cutime = 0;
            lru.tms_cstime = 0;
            prusage(&zru, &lru, pp->p_etime,
                    pp->p_btime);
#endif /* !BSDTIMES && !SEQUENT */

        }
#ifdef BACKPIPE
        pcond = ((tp == pp->p_friends && !inpipe) ||
                 (inpipe && pipehead->p_friends == tp && pp == pipetail));
#else  /* !BACKPIPE */
        pcond = (tp == pp->p_friends);
#endif /* BACKPIPE */
        if (pcond) {
            if (linp != linbuf)
                xputchar('\n');
            if (flag & SHELLDIR && !eq(tp->p_cwd->di_name, dcwd->di_name)) {
                xprintf(CGETS(17, 11, "(wd now: "));
                dtildepr(dcwd->di_name);
                xprintf(")\n");
            }
        }
#ifdef BACKPIPE
        if (inpipe) {
            /*
             * if pmaker == pipetail, we are finished that pipeline, and
             * can now skip to past the head
             */
            if (pmarker == pipetail) {
                inpipe = 0;
                pp = pipehead;
            }
            else {
            /*
             * set pp to one before the one we want next, so the while below
             * increments to the correct spot.
             */
                do
                    pp = pp->p_friends;
                while (pp->p_friends->p_friends != pmarker);
                pmarker = pp->p_friends;
            }
        }
        pcond = ((pp = pp->p_friends) != tp || inpipe);
#else /* !BACKPIPE */
        pcond = ((pp = pp->p_friends) != tp);
#endif /* BACKPIPE */
    } while (pcond);

    if (jobflags & PTIME && (jobflags & (PSTOPPED | PRUNNING)) == 0) {
        if (jobflags & NUMBER)
            xprintf("       ");
        ptprint(tp);
    }
    return (jobflags);
}

/*
 * All 4.3 BSD derived implementations are buggy and I've had enough.
 * The following implementation produces similar code and works in all
 * cases. The 4.3BSD one works only for <, >, !=
 */
# undef timercmp
#  define timercmp(tvp, uvp, cmp) \
      (((tvp)->tv_sec == (uvp)->tv_sec) ? \
           ((tvp)->tv_usec cmp (uvp)->tv_usec) : \
           ((tvp)->tv_sec  cmp (uvp)->tv_sec))

static void
ptprint(struct process *tp)
{
#ifdef BSDTIMES
    struct timeval tetime, diff;
    static struct timeval ztime;
    struct sysrusage ru;
    struct process *pp = tp;

    ru = zru;
    tetime = ztime;
    do {
        ruadd(&ru, &pp->p_rusage);
        tvsub(&diff, &pp->p_etime, &pp->p_btime);
        if (timercmp(&diff, &tetime, >))
            tetime = diff;
    } while ((pp = pp->p_friends) != tp);
    prusage(&zru, &ru, &tetime, &ztime);
#else /* !BSDTIMES */
# ifdef _SEQUENT_
    timeval_t tetime, diff;
    static timeval_t ztime;
    struct process_stats ru;
    struct process *pp = tp;

    ru = zru;
    tetime = ztime;
    do {
        ruadd(&ru, &pp->p_rusage);
        tvsub(&diff, &pp->p_etime, &pp->p_btime);
        if (timercmp(&diff, &tetime, >))
            tetime = diff;
    } while ((pp = pp->p_friends) != tp);
    prusage(&zru, &ru, &tetime, &ztime);
# else /* !_SEQUENT_ */
#  ifndef POSIX
    static time_t ztime = 0;
    static time_t zu_time = 0;
    static time_t zs_time = 0;
    time_t  tetime, diff;
    time_t  u_time, s_time;

#  else /* POSIX */
    static clock_t ztime = 0;
    static clock_t zu_time = 0;
    static clock_t zs_time = 0;
    clock_t tetime, diff;
    clock_t u_time, s_time;

#  endif /* POSIX */
    struct tms zts, rts;
    struct process *pp = tp;

    u_time = zu_time;
    s_time = zs_time;
    tetime = ztime;
    do {
        u_time += pp->p_utime;
        s_time += pp->p_stime;
        diff = pp->p_etime - pp->p_btime;
        if (diff > tetime)
            tetime = diff;
    } while ((pp = pp->p_friends) != tp);
    zts.tms_utime = zu_time;
    zts.tms_stime = zs_time;
    zts.tms_cutime = 0;
    zts.tms_cstime = 0;
    rts.tms_utime = u_time;
    rts.tms_stime = s_time;
    rts.tms_cutime = 0;
    rts.tms_cstime = 0;
    prusage(&zts, &rts, tetime, ztime);
# endif /* !_SEQUENT_ */
#endif  /* !BSDTIMES */
}

/*
 * dojobs - print all jobs
 */
/*ARGSUSED*/
void
dojobs(Char **v, struct command *c)
{
    struct process *pp;
    int flag = NUMBER | NAME | REASON;
    int     i;

    USE(c);
    if (chkstop)
        chkstop = 2;
    if (*++v) {
        if (v[1] || !eq(*v, STRml))
            stderror(ERR_JOBS);
        flag |= FANCY | JOBDIR;
    }
    for (i = 1; i <= pmaxindex; i++)
        for (pp = proclist.p_next; pp; pp = pp->p_next)
            if (pp->p_index == i && pp->p_procid == pp->p_jobid) {
                pp->p_flags &= ~PNEEDNOTE;
                if (!(pprint(pp, flag) & (PRUNNING | PSTOPPED)))
                    pflush(pp);
                break;
            }
}

/*
 * dofg - builtin - put the job into the foreground
 */
/*ARGSUSED*/
void
dofg(Char **v, struct command *c)
{
    struct process *pp;

    USE(c);
    okpcntl();
    ++v;
    do {
        pp = pfind(*v);
        if (!pstart(pp, 1)) {
            pp->p_procid = 0;
            stderror(ERR_NAME|ERR_BADJOB, pp->p_command, strerror(errno));
            continue;
        }
        pjwait(pp);
    } while (*v && *++v);
}

/*
 * %... - builtin - put the job into the foreground
 */
/*ARGSUSED*/
void
dofg1(Char **v, struct command *c)
{
    struct process *pp;

    USE(c);
    okpcntl();
    pp = pfind(v[0]);
    if (!pstart(pp, 1)) {
        pp->p_procid = 0;
        stderror(ERR_NAME|ERR_BADJOB, pp->p_command, strerror(errno));
        return;
    }
    pjwait(pp);
}

/*
 * dobg - builtin - put the job into the background
 */
/*ARGSUSED*/
void
dobg(Char **v, struct command *c)
{
    struct process *pp;

    USE(c);
    okpcntl();
    ++v;
    do {
        pp = pfind(*v);
        if (!pstart(pp, 0)) {
            pp->p_procid = 0;
            stderror(ERR_NAME|ERR_BADJOB, pp->p_command, strerror(errno));
        }
    } while (*v && *++v);
}

/*
 * %... & - builtin - put the job into the background
 */
/*ARGSUSED*/
void
dobg1(Char **v, struct command *c)
{
    struct process *pp;

    USE(c);
    pp = pfind(v[0]);
    if (!pstart(pp, 0)) {
        pp->p_procid = 0;
        stderror(ERR_NAME|ERR_BADJOB, pp->p_command, strerror(errno));
    }
}

/*
 * dostop - builtin - stop the job
 */
/*ARGSUSED*/
void
dostop(Char **v, struct command *c)
{
    USE(c);
#ifdef BSDJOBS
    pkill(++v, SIGSTOP);
#endif /* BSDJOBS */
}

/*
 * dokill - builtin - superset of kill (1)
 */
/*ARGSUSED*/
void
dokill(Char **v, struct command *c)
{
    int signum, len = 0;
    const char *name;
    Char *sigptr;

    USE(c);
    v++;
    if (v[0] && v[0][0] == '-') {
        if (v[0][1] == 'l') {
            for (signum = 0; signum <= nsig; signum++) {
                if ((name = mesg[signum].iname) != NULL) {
                    len += strlen(name) + 1;
                    if (len >= TermH - 1) {
                        xputchar('\n');
                        len = strlen(name) + 1;
                    }
                    xprintf("%s ", name);
                }
            }
            xputchar('\n');
            return;
        }
        sigptr = &v[0][1];
        if (v[0][1] == 's') {
            if (v[1]) {
                v++;
                sigptr = &v[0][0];
            } else {
                stderror(ERR_NAME | ERR_TOOFEW);
            }
        }
        if (Isdigit(*sigptr)) {
            char *ep;
            signum = strtoul(short2str(sigptr), &ep, 0);
            if (*ep || signum < 0 || signum > (MAXSIG-1))
                stderror(ERR_NAME | ERR_BADSIG);
        }
        else {
            for (signum = 0; signum <= nsig; signum++)
                if (mesg[signum].iname &&
                    eq(sigptr, str2short(mesg[signum].iname)))
                    goto gotsig;
            setname(short2str(sigptr));
            stderror(ERR_NAME | ERR_UNKSIG);
        }
gotsig:
        v++;
    }
    else
        signum = SIGTERM;
    pkill(v, signum);
}

static void
pkill(Char **v, int signum)
{
    struct process *pp, *np;
    int jobflags = 0, err1 = 0;
    pid_t     pid;
    Char *cp, **vp, **globbed;

    pchild_disabled++;
    cleanup_push(&pchild_disabled, disabled_cleanup);
    if (setintr) {
        pintr_disabled++;
        cleanup_push(&pintr_disabled, disabled_cleanup);
    }

    /* Avoid globbing %?x patterns */
    for (vp = v; vp && *vp; vp++)
        if (**vp == '%')
            (void) quote(*vp);

    v = glob_all_or_error(v);
    globbed = v;
    cleanup_push(globbed, blk_cleanup);


    while (v && (cp = *v)) {
        if (*cp == '%') {
            np = pp = pfind(cp);
            do
                jobflags |= np->p_flags;
            while ((np = np->p_friends) != pp);
#ifdef BSDJOBS
            switch (signum) {

            case SIGSTOP:
            case SIGTSTP:
            case SIGTTIN:
            case SIGTTOU:
                if ((jobflags & PRUNNING) == 0) {
# ifdef SUSPENDED
                    xprintf(CGETS(17, 12, "%S: Already suspended\n"), cp);
# else /* !SUSPENDED */
                    xprintf(CGETS(17, 13, "%S: Already stopped\n"), cp);
# endif /* !SUSPENDED */
                    err1++;
                    goto cont;
                }
                break;
                /*
                 * suspend a process, kill -CONT %, then type jobs; the shell
                 * says it is suspended, but it is running; thanks jaap..
                 */
            case SIGCONT:
                if (!pstart(pp, 0)) {
                    pp->p_procid = 0;
                    stderror(ERR_NAME|ERR_BADJOB, pp->p_command,
                             strerror(errno));
                }
                goto cont;
            default:
                break;
            }
#endif /* BSDJOBS */
            if (killpg(pp->p_jobid, signum) < 0) {
                xprintf("%S: %s\n", cp, strerror(errno));
                err1++;
            }
#ifdef BSDJOBS
            if (signum == SIGTERM || signum == SIGHUP)
                (void) killpg(pp->p_jobid, SIGCONT);
#endif /* BSDJOBS */
        }
        else if (!(Isdigit(*cp) || *cp == '-'))
            stderror(ERR_NAME | ERR_JOBARGS);
        else {
            char *ep;
#ifndef WINNT_NATIVE
            pid = strtol(short2str(cp), &ep, 10);
#else
            pid = strtoul(short2str(cp), &ep, 0);
#endif /* WINNT_NATIVE */
            if (*ep)
                stderror(ERR_NAME | ERR_JOBARGS);
            else if (kill(pid, signum) < 0) {
                xprintf("%d: %s\n", pid, strerror(errno));
                err1++;
                goto cont;
            }
#ifdef BSDJOBS
            if (signum == SIGTERM || signum == SIGHUP)
                (void) kill(pid, SIGCONT);
#endif /* BSDJOBS */
        }
cont:
        v++;
    }
    cleanup_until(&pchild_disabled);
    if (err1)
        stderror(ERR_SILENT);
}

/*
 * pstart - start the job in foreground/background
 */
int
pstart(struct process *pp, int foregnd)
{
    int rv = 0;
    struct process *np;
    /* We don't use jobflags in this function right now (see below) */
    /* long    jobflags = 0; */

    pchild_disabled++;
    cleanup_push(&pchild_disabled, disabled_cleanup);
    np = pp;
    do {
        /* We don't use jobflags in this function right now (see below) */
        /* jobflags |= np->p_flags; */
        if (np->p_flags & (PRUNNING | PSTOPPED)) {
            np->p_flags |= PRUNNING;
            np->p_flags &= ~PSTOPPED;
            if (foregnd)
                np->p_flags |= PFOREGND;
            else
                np->p_flags &= ~PFOREGND;
        }
    } while ((np = np->p_friends) != pp);
    if (!foregnd)
        pclrcurr(pp);
    (void) pprint(pp, foregnd ? NAME | JOBDIR : NUMBER | NAME | AMPERSAND);

    /* GrP run jobcmd hook if foregrounding */
    if (foregnd) {
        job_cmd(pp->p_command);
    }

#ifdef BSDJOBS
    if (foregnd) {
        rv = tcsetpgrp(FSHTTY, pp->p_jobid);
    }
    /*
     * 1. child process of csh (shell script) receives SIGTTIN/SIGTTOU
     * 2. parent process (csh) receives SIGCHLD
     * 3. The "csh" signal handling function pchild() is invoked
     *    with a SIGCHLD signal.
     * 4. pchild() calls wait3(WNOHANG) which returns 0.
     *    The child process is NOT ready to be waited for at this time.
     *    pchild() returns without picking-up the correct status
     *    for the child process which generated the SIGCHLD.
     * 5. CONSEQUENCE : csh is UNaware that the process is stopped
     * 6. THIS LINE HAS BEEN COMMENTED OUT : if (jobflags&PSTOPPED)
     *    (beto@aixwiz.austin.ibm.com - aug/03/91)
     * 7. I removed the line completely and added extra checks for
     *    pstart, so that if a job gets attached to and dies inside
     *    a debugger it does not confuse the shell. [christos]
     * 8. on the nec sx-4 there seems to be a problem, which requires
     *    a syscall(151, getpid(), getpid()) in osinit. Don't ask me
     *    what this is doing. [schott@rzg.mpg.de]
     */

    if (rv != -1)
        rv = killpg(pp->p_jobid, SIGCONT);
#endif /* BSDJOBS */
    cleanup_until(&pchild_disabled);
    return rv != -1;
}

void
panystop(int neednl)
{
    struct process *pp;

    chkstop = 2;
    for (pp = proclist.p_next; pp; pp = pp->p_next)
        if (pp->p_flags & PSTOPPED)
            stderror(ERR_STOPPED, neednl ? "\n" : "");
}

struct process *
pfind(Char *cp)
{
    struct process *pp, *np;

    if (cp == 0 || cp[1] == 0 || eq(cp, STRcent2) || eq(cp, STRcentplus)) {
        if (pcurrent == NULL)
            stderror(ERR_NAME | ERR_JOBCUR);
        return (pcurrent);
    }
    if (eq(cp, STRcentminus) || eq(cp, STRcenthash)) {
        if (pprevious == NULL)
            stderror(ERR_NAME | ERR_JOBPREV);
        return (pprevious);
    }
    if (Isdigit(cp[1])) {
        int     idx = atoi(short2str(cp + 1));

        for (pp = proclist.p_next; pp; pp = pp->p_next)
            if (pp->p_index == idx && pp->p_procid == pp->p_jobid)
                return (pp);
        stderror(ERR_NAME | ERR_NOSUCHJOB);
    }
    np = NULL;
    for (pp = proclist.p_next; pp; pp = pp->p_next)
        if (pp->p_procid == pp->p_jobid) {
            if (cp[1] == '?') {
                Char *dp;

                for (dp = pp->p_command; *dp; dp++) {
                    if (*dp != cp[2])
                        continue;
                    if (prefix(cp + 2, dp))
                        goto match;
                }
            }
            else if (prefix(cp + 1, pp->p_command)) {
        match:
                if (np)
                    stderror(ERR_NAME | ERR_AMBIG);
                np = pp;
            }
        }
    if (np)
        return (np);
    stderror(ERR_NAME | (cp[1] == '?' ? ERR_JOBPAT : ERR_NOSUCHJOB));
    /* NOTREACHED */
    return (0);
}


/*
 * pgetcurr - find most recent job that is not pp, preferably stopped
 */
static struct process *
pgetcurr(struct process *pp)
{
    struct process *np;
    struct process *xp = NULL;

    for (np = proclist.p_next; np; np = np->p_next)
        if (np != pcurrent && np != pp && np->p_procid &&
            np->p_procid == np->p_jobid) {
            if (np->p_flags & PSTOPPED)
                return (np);
            if (xp == NULL)
                xp = np;
        }
    return (xp);
}

/*
 * donotify - flag the job so as to report termination asynchronously
 */
/*ARGSUSED*/
void
donotify(Char **v, struct command *c)
{
    struct process *pp;

    USE(c);
    pp = pfind(*++v);
    pp->p_flags |= PNOTIFY;
}

#ifdef SIGSYNCH
static void
synch_handler(int sno)
{
    USE(sno);
}
#endif /* SIGSYNCH */

/*
 * Do the fork and whatever should be done in the child side that
 * should not be done if we are not forking at all (like for simple builtin's)
 * Also do everything that needs any signals fiddled with in the parent side
 *
 * Wanttty tells whether process and/or tty pgrps are to be manipulated:
 *      -1:     leave tty alone; inherit pgrp from parent
 *       0:     already have tty; manipulate process pgrps only
 *       1:     want to claim tty; manipulate process and tty pgrps
 * It is usually just the value of tpgrp.
 */

pid_t
pfork(struct command *t, int wanttty)
{
    pid_t pid;
    int    ignint = 0;
    pid_t pgrp;
#ifdef SIGSYNCH
    struct sigaction osa, nsa;
#endif /* SIGSYNCH */

    /*
     * A child will be uninterruptible only under very special conditions.
     * Remember that the semantics of '&' is implemented by disconnecting the
     * process from the tty so signals do not need to ignored just for '&'.
     * Thus signals are set to default action for children unless: we have had
     * an "onintr -" (then specifically ignored) we are not playing with
     * signals (inherit action)
     */
    if (setintr)
        ignint = (tpgrp == -1 && (t->t_dflg & F_NOINTERRUPT))
            || (gointr && eq(gointr, STRminus));

    /*
     * Check for maximum nesting of 16 processes to avoid Forking loops
     */
    if (child == 16)
        stderror(ERR_NESTING, 16);
#ifdef SIGSYNCH
    nsa.sa_handler = synch_handler;
    sigfillset(&nsa.sa_mask);
    nsa.sa_flags = SA_RESTART;
    if (sigaction(SIGSYNCH, &nsa, &osa))
        stderror(ERR_SYSTEM, "pfork: sigaction set", strerror(errno));
#endif /* SIGSYNCH */
    /*
     * Hold pchild() until we have the process installed in our table.
     */
    if (wanttty < 0) {
        pchild_disabled++;
        cleanup_push(&pchild_disabled, disabled_cleanup);
    }
    while ((pid = fork()) == -1)
        if (setintr == 0)
            (void) sleep(FORKSLEEP);
        else
            stderror(ERR_NOPROC);
    if (pid == 0) {
        (void)cleanup_push_mark(); /* Never to be popped */
        pchild_disabled = 0;
        settimes();
        pgrp = pcurrjob ? pcurrjob->p_jobid : getpid();
        pflushall();
        pcurrjob = NULL;
#if !defined(BSDTIMES) && !defined(_SEQUENT_) 
        timesdone = 0;
#endif /* !defined(BSDTIMES) && !defined(_SEQUENT_) */
        child++;
        if (setintr) {
            setintr = 0;        /* until I think otherwise */
            /*
             * Children just get blown away on SIGINT, SIGQUIT unless "onintr
             * -" seen.
             */
            (void) signal(SIGINT, ignint ? SIG_IGN : SIG_DFL);
            (void) signal(SIGQUIT, ignint ? SIG_IGN : SIG_DFL);
#ifdef BSDJOBS
            if (wanttty >= 0) {
                /* make stoppable */
                (void) signal(SIGTSTP, SIG_DFL);
                (void) signal(SIGTTIN, SIG_DFL);
                (void) signal(SIGTTOU, SIG_DFL);
            }
#endif /* BSDJOBS */
            sigaction(SIGTERM, &parterm, NULL);
        }
        else if (tpgrp == -1 && (t->t_dflg & F_NOINTERRUPT)) {
            (void) signal(SIGINT, SIG_IGN);
            (void) signal(SIGQUIT, SIG_IGN);
        }
#ifdef OREO
        signal(SIGIO, SIG_IGN); /* ignore SIGIO in child too */
#endif /* OREO */

        pgetty(wanttty, pgrp);
        /*
         * Nohup and nice apply only to NODE_COMMAND's but it would be nice
         * (?!?) if you could say "nohup (foo;bar)" Then the parser would have
         * to know about nice/nohup/time
         */
        if (t->t_dflg & F_NOHUP)
            (void) signal(SIGHUP, SIG_IGN);
        if (t->t_dflg & F_NICE) {
            int nval = SIGN_EXTEND_CHAR(t->t_nice);
#ifdef HAVE_SETPRIORITY
            if (setpriority(PRIO_PROCESS, 0, nval) == -1 && errno)
                stderror(ERR_SYSTEM, "setpriority", strerror(errno));
#else /* !HAVE_SETPRIORITY */
            (void) nice(nval);
#endif /* !HAVE_SETPRIORITY */
        }
#ifdef F_VER
        if (t->t_dflg & F_VER) {
            tsetenv(STRSYSTYPE, t->t_systype ? STRbsd43 : STRsys53);
            dohash(NULL, NULL);
        }
#endif /* F_VER */
#ifdef SIGSYNCH
        /* rfw 8/89 now parent can continue */
        if (kill(getppid(), SIGSYNCH))
            stderror(ERR_SYSTEM, "pfork child: kill", strerror(errno));
#endif /* SIGSYNCH */

    }
    else {
#ifdef POSIXJOBS
        if (wanttty >= 0) {
            /*
             * `Walking' process group fix from Beto Appleton.
             * (beto@aixwiz.austin.ibm.com)
             * If setpgid fails at this point that means that
             * our process leader has died. We flush the current
             * job and become the process leader ourselves.
             * The parent will figure that out later.
             */
            pgrp = pcurrjob ? pcurrjob->p_jobid : pid;
            if (setpgid(pid, pgrp) == -1 && errno == EPERM) {
                pcurrjob = NULL;
                /* 
                 * We don't care if this causes an error here;
                 * then we are already in the right process group
                 */
                (void) setpgid(pid, pgrp = pid);
            }
        }
#endif /* POSIXJOBS */
        palloc(pid, t);
#ifdef SIGSYNCH
        {
            sigset_t pause_mask;

        /*
         * rfw 8/89 Wait for child to own terminal.  Solves half of ugly
         * synchronization problem.  With this change, we know that the only
         * reason setpgrp to a previous process in a pipeline can fail is that
         * the previous process has already exited. Without this hack, he may
         * either have exited or not yet started to run.  Two uglies become
         * one.
         */
            sigprocmask(SIG_BLOCK, NULL, &pause);
            sigdelset(&pause_mask, SIGCHLD);
            sigdelset(&pause_mask, SIGSYNCH);
            sigsuspend(&pause_mask);
            handle_pending_signals();
            if (sigaction(SIGSYNCH, &osa, NULL))
                stderror(ERR_SYSTEM, "pfork parent: sigaction restore",
                         strerror(errno));
        }
#endif /* SIGSYNCH */

        if (wanttty < 0)
            cleanup_until(&pchild_disabled);
    }
    return (pid);
}

static void
okpcntl(void)
{
    if (tpgrp == -1)
        stderror(ERR_JOBCONTROL);
    if (tpgrp == 0)
        stderror(ERR_JOBCTRLSUB);
}


static void
setttypgrp(int pgrp)
{
    /*
     * If we are piping out a builtin, eg. 'echo | more' things can go
     * out of sequence, i.e. the more can run before the echo. This
     * can happen even if we have vfork, since the echo will be forked
     * with the regular fork. In this case, we need to set the tty
     * pgrp ourselves. If that happens, then the process will be still
     * alive. And the tty process group will already be set.
     * This should fix the famous sequent problem as a side effect:
     *    The controlling terminal is lost if all processes in the
     *    terminal process group are zombies. In this case tcgetpgrp()
     *    returns 0. If this happens we must set the terminal process
     *    group again.
     */
    if (tcgetpgrp(FSHTTY) != pgrp) {
#ifdef POSIXJOBS
        struct sigaction old;

        /*
         * tcsetpgrp will set SIGTTOU to all the the processes in 
         * the background according to POSIX... We ignore this here.
         */
        sigaction(SIGTTOU, NULL, &old);
        signal(SIGTTOU, SIG_IGN);
#endif
        (void) tcsetpgrp(FSHTTY, pgrp);
# ifdef POSIXJOBS
        sigaction(SIGTTOU, &old, NULL);
# endif

    }
}


/*
 * if we don't have vfork(), things can still go in the wrong order
 * resulting in the famous 'Stopped (tty output)'. But some systems
 * don't permit the setpgid() call, (these are more recent secure
 * systems such as ibm's aix), when they do. Then we'd rather print 
 * an error message than hang the shell!
 * I am open to suggestions how to fix that.
 */
void
pgetty(int wanttty, pid_t pgrp)
{
#ifdef BSDJOBS
# ifdef POSIXJOBS
    sigset_t oset, set;
# endif /* POSIXJOBS */

    jobdebug_xprintf(("wanttty %d pid %d opgrp%d pgrp %d tpgrp %d\n",
                      wanttty, (int)getpid(), (int)pgrp, (int)mygetpgrp(),
                      (int)tcgetpgrp(FSHTTY)));
# ifdef POSIXJOBS
    /*
     * christos: I am blocking the tty signals till I've set things
     * correctly....
     */
    if (wanttty > 0) {
        sigemptyset(&set);
        sigaddset(&set, SIGTSTP);
        sigaddset(&set, SIGTTIN);
        (void)sigprocmask(SIG_BLOCK, &set, &oset);
        cleanup_push(&oset, sigprocmask_cleanup);
    }
# endif /* POSIXJOBS */

# ifndef POSIXJOBS
    if (wanttty > 0)
        setttypgrp(pgrp);
# endif /* !POSIXJOBS */

    /*
     * From: Michael Schroeder <mlschroe@immd4.informatik.uni-erlangen.de>
     * Don't check for tpgrp >= 0 so even non-interactive shells give
     * background jobs process groups Same for the comparison in the other part
     * of the #ifdef
     */
    if (wanttty >= 0) {
        if (setpgid(0, pgrp) == -1) {
# ifdef POSIXJOBS
            /* Walking process group fix; see above */
            if (setpgid(0, pgrp = getpid()) == -1) {
# endif /* POSIXJOBS */
                stderror(ERR_SYSTEM, "setpgid child:\n", strerror(errno));
                xexit(0);
# ifdef POSIXJOBS
            }
            wanttty = pgrp;  /* Now we really want the tty, since we became the
                              * the process group leader
                              */
# endif /* POSIXJOBS */
        }
    }

# ifdef POSIXJOBS
    if (wanttty > 0) {
        setttypgrp(pgrp);
        cleanup_until(&oset);
    }
# endif /* POSIXJOBS */

    jobdebug_xprintf(("wanttty %d pid %d pgrp %d tpgrp %d\n",
                      wanttty, getpid(), mygetpgrp(), tcgetpgrp(FSHTTY)));

    if (tpgrp > 0)
        tpgrp = 0;              /* gave tty away */
#endif /* BSDJOBS */
}