feat(process): fork/exec/id/session + system/popen

include/unistd.h

@@ -15,6 +15,15 @@
  *   Level 2 (muslmimic): dup3, pipe2 (Linux extensions), truncate (XSI),
  *                        lseek64 (glibc LFS alias of lseek on x86_64).
  *
+ * Process control (todo 20) is declared further down:
+ *
+ *   Level 1 (onlyposix): fork, exec family (execl/execlp/execle/execv/
+ *                        execvp/execve/fexecve), getpid, getppid, getuid,
+ *                        geteuid, getgid, getegid, setuid, seteuid, setgid,
+ *                        setegid, getgroups, setpgid, getpgrp, setsid.
+ *   Level 2 (muslmimic): vfork, setpgrp (obsolescent), setgroups, getpgid,
+ *                        getsid, tcgetpgrp, tcsetpgrp (XSI).
+ *
  * The open-flag constants (O_RDONLY, O_CREAT, O_CLOEXEC, ...) belong to
  * <fcntl.h> and are deliberately not defined here; the oflag arguments below
  * are plain int and take their values from that header. The optional mode
@@ -23,7 +32,10 @@
  *
  * None of these declarations carry an intent attribute: every function
  * performs I/O with side effects and reports failures through errno, so
- * const/pure would be unsound.
+ * const/pure would be unsound. The process functions additionally must not
+ * be const/pure because their results are process state — marking getpid
+ * const, for example, would let the compiler hoist it across fork() and
+ * observe the parent's pid inside the child.
  */
 
 #include <vlibc/features.h>
@@ -174,8 +186,119 @@ access(const char *path, int amode);
 int
 faccessat(int fd, const char *path, int amode, int flag);
 
+/* Level 1 (POSIX base): process control. */
+
+/*
+ * Create a child process that is a copy of the caller: fork() returns 0
+ * in the child, the child's pid in the parent, and -1 with errno set on
+ * failure.
+ *
+ * Around the fork, the atfork hook table runs: the prepare handlers in
+ * reverse registration order before the fork, then the child handlers in
+ * the child and the parent handlers in the parent, both in registration
+ * order (POSIX pthread_atfork protocol). The table is registered by
+ * pthread_atfork() (todo 45) and is empty — a no-op — in a process
+ * without threads.
+ */
+pid_t
+fork(void);
+
+/*
+ * Replace the calling process image. Success never returns; -1 with
+ * errno set otherwise. execl/execle/execlp take the arguments as a
+ * varargs list terminated by (char *)NULL (execle is followed by one
+ * final char *const envp[] argument); execv/execvp take an argv array.
+ * execvp and execlp search the PATH environment variable (default
+ * "/bin:/usr/bin") when file contains no '/'. execv/execvp/execl/execlp
+ * use the caller's environment; execve/execle/fexecve take envp.
+ */
+// NOLINTBEGIN(bugprone-easily-swappable-parameters)
+int
+execl(const char *path, const char *arg0, ...);
+
+int
+execle(const char *path, const char *arg0, ...);
+
+int
+execlp(const char *file, const char *arg0, ...);
+// NOLINTEND(bugprone-easily-swappable-parameters)
+
+int
+execv(const char *path, char *const argv[]);
+
+int
+execvp(const char *file, char *const argv[]);
+
+int
+execve(const char *path, char *const argv[], char *const envp[]);
+
+/*
+ * Like execve, but the image is the open descriptor fd.
+ */
+int
+fexecve(int fd, char *const argv[], char *const envp[]);
+
+/*
+ * Process ids. getpid/getppid/getuid/geteuid/getgid/getegid read the
+ * kernel's per-process ids; setuid/seteuid/setgid/setegid change them.
+ * seteuid/setegid change only the effective id (setresuid/setresgid
+ * underneath: Linux has no seteuid/setegid syscall).
+ */
+pid_t
+getpid(void);
+
+pid_t
+getppid(void);
+
+uid_t
+getuid(void);
+
+uid_t
+geteuid(void);
+
+gid_t
+getgid(void);
+
+gid_t
+getegid(void);
+
+int
+setuid(uid_t uid);
+
+int
+seteuid(uid_t euid);
+
+int
+setgid(gid_t gid);
+
+int
+setegid(gid_t egid);
+
+/*
+ * Supplementary groups: getgroups fills grouplist with up to gidsetsize
+ * group ids and returns the total count (0 can be passed to count only);
+ * setgroups (level 2, XSI) installs the list.
+ */
+int
+getgroups(int gidsetsize, gid_t grouplist[]);
+
+/*
+ * Session control: setpgid moves pid into process group pgid (0 means
+ * the caller / the caller's pid); getpgrp returns the caller's process
+ * group; setsid creates a new session with the caller as leader and
+ * returns the new session id.
+ */
+int
+setpgid(pid_t pid, pid_t pgid);
+
+pid_t
+getpgrp(void);
+
+pid_t
+setsid(void);
+
 #if VLIBC_LEVEL_GE(2)
-/* Level 2 (muslmimic): Linux extensions + XSI. */
+/* Level 2 (muslmimic): Linux extensions + XSI + obsolescent. */
 
 /*
  * Like dup2(), but with descriptor flags (O_CLOEXEC from <fcntl.h>) applied
@@ -205,6 +328,48 @@ truncate(const char *path, off_t length);
  */
 off_t
 lseek64(int fildes, off_t offset, int whence);
+
+/*
+ * vfork (obsolescent): like fork(), but the child borrows the parent's
+ * address space until it execs or exits.
+ */
+pid_t
+vfork(void);
+
+/*
+ * setpgrp (obsolescent): setpgid(0, 0) — the caller becomes the leader
+ * of its own process group.
+ */
+int
+setpgrp(void);
+
+/*
+ * setgroups (XSI): install the supplementary group list of gidsetsize
+ * entries. Requires privilege.
+ */
+int
+setgroups(size_t gidsetsize, const gid_t *grouplist);
+
+/*
+ * getpgid/getsid (XSI): the process group / session of pid (0 = the
+ * caller).
+ */
+pid_t
+getpgid(pid_t pid);
+
+pid_t
+getsid(pid_t pid);
+
+/*
+ * Controlling-terminal foreground process group (XSI): tcgetpgrp reads
+ * the foreground group of the terminal on fildes; tcsetpgrp makes pgid
+ * the foreground group.
+ */
+pid_t
+tcgetpgrp(int fildes);
+
+int
+tcsetpgrp(int fildes, pid_t pgid);
 #endif /* VLIBC_LEVEL_GE(2) */
 
 #ifdef __cplusplus