Switch b_alias() from optget() to getopt_long()

Another step in switching from the DocOpt base option processing done by
the AST `optget()` function to the borg standard `getopt_long()`.

It also introduces a system-wide ksh config script to enable auto-loading
functions included with ksh by default. This allows the use of the
auto-loaded `_ksh_print_help` function. As well as the dirs/popd/pushd
set of functions that have been the sole functions included with ksh
forever but not automatically enabled.

Related #507
Fixes #835

src/cmd/ksh93/bltins/alias.c

@@ -19,6 +19,8 @@
  ***********************************************************************/
 #include "config_ast.h"  // IWYU pragma: keep
 
+#include <getopt.h>
+#include <stdlib.h>
 #include <string.h>
 
 #include "argnod.h"
@@ -27,31 +29,38 @@
 #include "defs.h"
 #include "error.h"
 #include "name.h"
-#include "option.h"
 #include "shcmd.h"
 #include "variables.h"
 
+static const char *short_options = ":ptx";
+static const struct option long_options[] = {{"help", 0, NULL, 1},  // all builtins supports --help
+                                             {NULL, 0, NULL, 0}};
+
 int b_alias(int argc, char *argv[], Shbltin_t *context) {
     UNUSED(argc);
-    nvflag_t nvflags = NV_NOARRAY | NV_NOSCOPE | NV_ASSIGN;
+    int opt;
     Dt_t *troot;
-    int n;
     struct tdata tdata;
+    Shell_t *shp = context->shp;
+    char *cmd = argv[0];
+    nvflag_t nvflags = NV_NOARRAY | NV_NOSCOPE | NV_ASSIGN;
 
     memset(&tdata, 0, sizeof(tdata));
-    tdata.sh = context->shp;
+    tdata.sh = shp;
     troot = tdata.sh->alias_tree;
-    if (*argv[0] == 'h') nvflags = NV_TAGGED;
-    if (sh_isoption(tdata.sh, SH_BASH)) tdata.prefix = argv[0];
+    if (sh_isoption(tdata.sh, SH_BASH)) tdata.prefix = cmd;
     if (!argv[1]) return setall(argv, nvflags, troot, &tdata);
 
-    opt_info.offset = 0;
-    opt_info.index = 1;
-    *opt_info.option = 0;
     tdata.argnum = 0;
     tdata.aflag = *argv[1];
-    while ((n = optget(argv, sh_optalias))) {
-        switch (n) {  //!OCLINT(MissingDefaultStatement)
+
+    optind = 0;
+    while ((opt = getopt_long(argc, argv, short_options, long_options, NULL)) != -1) {
+        switch (opt) {
+            case 1: {
+                builtin_print_help(shp, cmd);
+                return 0;
+            }
             case 'p': {
                 tdata.prefix = argv[0];
                 break;
@@ -65,21 +74,18 @@ int b_alias(int argc, char *argv[], Shbltin_t *context) {
                 break;
             }
             case ':': {
-                errormsg(SH_DICT, 2, "%s", opt_info.arg);
-                break;
+                builtin_missing_argument(shp, cmd, argv[optind - 1]);
+                return 2;
             }
             case '?': {
-                errormsg(SH_DICT, ERROR_usage(0), "%s", opt_info.arg);
+                builtin_unknown_option(shp, cmd, argv[optind - 1]);
                 return 2;
             }
+            default: { abort(); }
         }
     }
-    if (error_info.errors) {
-        errormsg(SH_DICT, ERROR_usage(2), "%s", optusage(NULL));
-        __builtin_unreachable();
-    }
 
-    argv += (opt_info.index - 1);
+    argv += (optind - 1);
     if (!nv_isflag(nvflags, NV_TAGGED)) return setall(argv, nvflags, troot, &tdata);
 
     // Hacks to handle hash -r | --.
@@ -99,6 +105,7 @@ int b_alias(int argc, char *argv[], Shbltin_t *context) {
             }
         }
     }
+
     troot = tdata.sh->track_tree;
     return setall(argv, nvflags, troot, &tdata);
 }

src/cmd/ksh93/data/builtins.c

@@ -29,6 +29,8 @@
 
 #include "builtins.h"
 #include "jobs.h"
+#include "sfio.h"
+
 #define bltin(x) (b_##x)
 
 // In the last beta release that came out from AT&T, all the builtins for standard commands
@@ -228,3 +230,28 @@ const char e_nofork[] = "cannot fork";
 const char e_nosignal[] = "%s: unknown signal name";
 const char e_condition[] = "condition(s) required";
 const char e_cneedsarg[] = "-c requires argument";
+
+// Error message for missing flag argument.
+#define BUILTIN_ERR_MISSING "\n%s: Expected argument for flag '%s'\n"
+// Error message for unrecognized flag.
+#define BUILTIN_ERR_UNKNOWN "\n%s: Unknown flag '%s'\n"
+
+// Invoke a helper function or command to print a subset of the man page for the command. Such as
+// from doing "cmd --help".
+void builtin_print_help(Shell_t *shp, const char *cmd) {
+    const char *argv[3] = {"_ksh_print_help", cmd, NULL};
+    sh_eval(shp, sh_sfeval(argv), 0);
+    return;
+}
+
+// Error reporting for encounter with unknown option when parsing command arguments.
+void builtin_unknown_option(Shell_t *shp, const char *cmd, const char *opt) {
+    builtin_print_help(shp, cmd);
+    sfprintf(sfstderr, BUILTIN_ERR_UNKNOWN, cmd, opt);
+}
+
+// Error reporting for encounter with missing argument when parsing command arguments.
+void builtin_missing_argument(Shell_t *shp, const char *cmd, const char *opt) {
+    builtin_print_help(shp, cmd);
+    sfprintf(sfstderr, BUILTIN_ERR_MISSING, cmd, opt);
+}

src/cmd/ksh93/data/config.sh

@@ -0,0 +1,23 @@
+#
+# This script is sourced first when the shell starts. It's purpose is to perform basic setup of the
+# shell state. For example, where to find autoloaded functions that ship with the shell and arrange
+# for them to be loaded on demand. It should not do anything not suitable for every single new ksh
+# process. In particular it should be as fast as possible.
+#
+
+# Arrange for standard autoloaded functions to be available. The test for whether or not FPATH is
+# already set isn't technically necessary since empty path components are guaranteed not to be
+# equivalent to `.` (the CWD). But I prefer to be paranoid since doing so is cheap.
+__fpath="${.sh.install_prefix}/share/ksh/functions"
+if [[ -z $FPATH ]]
+then
+    FPATH="$__fpath"
+else
+    FPATH="$__fpath:$FPATH"
+fi
+
+for f in "$__fpath"/*
+do
+    typeset -fu $(basename $f)
+done
+unset __fpath

src/cmd/ksh93/functions/_ksh_print_help

@@ -0,0 +1,43 @@
+#
+# Given a ksh man page emit the text of the SYNOPSIS and FLAGS sections and nothing else. This is
+# primarily meant to be used by builtins (either special or normal commands) to emit a subset of its
+# man page when invoked with `cmd --help` or the argument parsing for the command detects an error.
+# Such as via the `builtin_print_help()` function in the ksh source.
+#
+function _ksh_print_help {
+    local cmd=$1
+    local fname="${.sh.install_prefix}/share/ksh/man/man1/${cmd}.1"
+
+    [[ -f $fname ]] || return 1  # can't find the man page; give up
+
+    # Find the `ul` command to translate x\cHx sequences, etc., into bold chars and such. Pagers
+    # like `less` do this but better to not rely on them. If `ul` is not available just pipe the
+    # nroff output unmodified.
+    local ul=$(whence ul)
+    [[ -z $ul ]] && ul=cat
+
+    # Try to determine an optimal line length for the output.
+    local cols=80
+    if [[ -t 0 && -t 1 ]]
+    then
+        stty size | read _ cols
+    fi
+    cols=$(( cols - 4 ))  # leave a little space on the right side of the screen
+
+    # Convert the nroff source for the command into something that looks good on the terminal.
+    # Specifically, by elimating all the text outside of the SYNOPSIS and FLAGS sections while
+    # attempting to preserve some of the highlighting understood by the terminal.
+    nroff -c -man -rLL=${cols}n "$fname" |
+        sed -n \
+            -e '/^S.*Y.*N.*O.*P.*S.*I.*S$/,/^[^ ]/ {
+                    /^S/p
+                    /^[^ ]/b
+                    p
+                }' \
+            -e '/^F.*L.*A.*G.*S$/,/^[^ ]/ {
+                    /^F/p
+                    /^[^ ]/b
+                    p
+                }' |
+        uniq | "$ul"
+}

src/cmd/ksh93/include/builtins.h

@@ -68,6 +68,9 @@ struct tdata {
 
 extern int setall(char **, nvflag_t, Dt_t *, struct tdata *);
 extern void print_scan(Sfio_t *file, nvflag_t flag, Dt_t *root, bool omit_attrs, struct tdata *tp);
+extern void builtin_print_help(Shell_t *shp, const char *cmd);
+extern void builtin_unknown_option(Shell_t *shp, const char *cmd, const char *opt);
+extern void builtin_missing_argument(Shell_t *shp, const char *cmd, const char *opt);
 
 // Entry points for shell special builtins.
 extern int b_alias(int, char *[], Shbltin_t *);
@@ -160,7 +163,6 @@ extern const char e_defined[];
 
 // For option parsing.
 extern const char sh_set[];
-extern const char sh_optalias[];
 extern const char sh_optbreak[];
 extern const char sh_optbuiltin[];
 extern const char sh_optcd[];

src/cmd/ksh93/meson.build

@@ -39,6 +39,9 @@ shcomp_exe = executable('shcomp', ['sh/shcomp.c'], c_args: shared_c_args,
     dependencies: [libm_dep, libexecinfo_dep, libdl_dep, libsocket_dep, libnsl_dep],
     install: true)
 
+install_subdir('functions', install_dir: 'share/ksh')
+install_data(['data/config.sh'], install_dir: 'share/ksh')
+
 test_dir = join_paths(meson.current_source_dir(), 'tests')
 test_driver = join_paths(test_dir, 'util', 'run_test.sh')
 

src/cmd/ksh93/sh/main.c

@@ -162,6 +162,7 @@ int sh_main(int ac, char *av[], Shinit_f userinit) {
             sh_onoption(shp, SH_MONITOR);
         }
         job_init(shp, sh_isoption(shp, SH_LOGIN_SHELL));
+        sh_source(shp, iop, INSTALL_PREFIX "/share/ksh/config.sh");
         if (sh_isoption(shp, SH_LOGIN_SHELL)) {
             //  System profile.
             sh_source(shp, iop, e_sysprofile);