]> git.cameronkatri.com Git - mandoc.git/blobdiff - cgi.c
fix unchecked snprintf(3) in page header printing:
[mandoc.git] / cgi.c
diff --git a/cgi.c b/cgi.c
index 67ad47bb939b1e553e9a613d72f3eb04233ff938..35b20160690920241a7ec947fae8e2de598b5edd 100644 (file)
--- a/cgi.c
+++ b/cgi.c
-/*     $Id: cgi.c,v 1.3 2011/11/13 10:12:05 schwarze Exp $ */
+/*     $Id: cgi.c,v 1.50 2014/03/23 12:19:03 schwarze Exp $ */
+/*
+ * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <sys/wait.h>
+
 #include <assert.h>
+#include <ctype.h>
+#include <errno.h>
+#include <dirent.h>
 #include <fcntl.h>
+#include <limits.h>
 #include <regex.h>
 #include <stdio.h>
 #include <stdarg.h>
+#include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
+#include <unistd.h>
+
+#if defined(__sun)
+/* for stat() */
+#include <fcntl.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#endif
 
 #include "apropos_db.h"
 #include "mandoc.h"
+#include "mandoc_aux.h"
+#include "mdoc.h"
+#include "man.h"
+#include "main.h"
+#include "manpath.h"
+#include "mandocdb.h"
+
+#if defined(__linux__) || defined(__sun)
+# include <db_185.h>
+#else
+# include <db.h>
+#endif
 
-/*
- * The page a request is trying to make.
- */
 enum   page {
        PAGE_INDEX,
        PAGE_SEARCH,
+       PAGE_SHOW,
        PAGE__MAX
 };
 
-/*
- * Key-value pair.  
- * Both key and val are on the heap.
- */
-struct kval {
-       char            *key;
-       char            *val;
+struct paths {
+       char            *name;
+       char            *path;
 };
 
 /*
- * The media type, determined by suffix, of the requesting or responding
- * context.
+ * A query as passed to the search function.
  */
-enum   media {
-       MEDIA_HTML,
-       MEDIA__MAX
+struct query {
+       const char      *arch; /* architecture */
+       const char      *sec; /* manual section */
+       const char      *expr; /* unparsed expression string */
+       int              manroot; /* manroot index (or -1)*/
+       int              legacy; /* whether legacy mode */
 };
 
-/*
- * An HTTP request.
- */
 struct req {
-       struct kval     *fields; /* query fields */
-       size_t           fieldsz;
-       enum media       media;
+       struct query     q;
+       struct paths    *p;
+       size_t           psz;
        enum page        page;
 };
 
-#if 0
-static void             html_printtext(const char *);
-#endif
-static int              kval_decode(char *);
-static void             kval_parse(struct kval **, size_t *, char *);
-static void             kval_free(struct kval *, size_t);
+static int              atou(const char *, unsigned *);
+static void             catman(const struct req *, const char *);
+static int              cmp(const void *, const void *);
+static void             format(const struct req *, const char *);
+static void             html_print(const char *);
+static void             html_printquery(const struct req *);
+static void             html_putchar(char);
+static int              http_decode(char *);
+static void             http_parse(struct req *, char *);
+static void             http_print(const char *);
+static void             http_putchar(char);
+static void             http_printquery(const struct req *);
+static int              pathstop(DIR *);
+static void             pathgen(DIR *, char *, struct req *);
 static void             pg_index(const struct req *, char *);
 static void             pg_search(const struct req *, char *);
-#if 0
-static void             pg_searchres(struct rec *, size_t, void *);
-#endif
+static void             pg_show(const struct req *, char *);
+static void             resp_bad(void);
+static void             resp_baddb(void);
+static void             resp_error400(void);
+static void             resp_error404(const char *);
+static void             resp_begin_html(int, const char *);
+static void             resp_begin_http(int, const char *);
+static void             resp_end_html(void);
+static void             resp_index(const struct req *);
+static void             resp_search(struct res *, size_t, void *);
+static void             resp_searchform(const struct req *);
+
+static const char       *progname; /* cgi script name */
+static const char       *cache; /* cache directory */
+static const char       *css; /* css directory */
+static const char       *host; /* hostname */
 
 static const char * const pages[PAGE__MAX] = {
        "index", /* PAGE_INDEX */ 
        "search", /* PAGE_SEARCH */
+       "show", /* PAGE_SHOW */
 };
 
-static const char * const medias[MEDIA__MAX] = {
-       "html", /* MEDIA_HTML */
-};
+/*
+ * This is just OpenBSD's strtol(3) suggestion.
+ * I use it instead of strtonum(3) for portability's sake.
+ */
+static int
+atou(const char *buf, unsigned *v)
+{
+       char            *ep;
+       long             lval;
+
+       errno = 0;
+       lval = strtol(buf, &ep, 10);
+       if (buf[0] == '\0' || *ep != '\0')
+               return(0);
+       if ((errno == ERANGE && (lval == LONG_MAX || 
+                                       lval == LONG_MIN)) ||
+                       (lval > INT_MAX || lval < 0))
+               return(0);
+
+       *v = (unsigned int)lval;
+       return(1);
+}
 
-#if 0
+/*
+ * Print a character, escaping HTML along the way.
+ * This will pass non-ASCII straight to output: be warned!
+ */
 static void
-html_printtext(const char *p)
+html_putchar(char c)
 {
-       char             c;
 
-       while ('\0' != *p)
-               switch ((c = *p++)) {
-               case ('"'):
-                       printf("&quote;");
-                       break;
-               case ('&'):
-                       printf("&amp;");
-                       break;
-               case ('>'):
-                       printf("&gt;");
-                       break;
-               case ('<'):
-                       printf("&lt;");
-                       break;
-               default:
-                       putchar((unsigned char)c);
-                       break;
-               }
+       switch (c) {
+       case ('"'):
+               printf("&quote;");
+               break;
+       case ('&'):
+               printf("&amp;");
+               break;
+       case ('>'):
+               printf("&gt;");
+               break;
+       case ('<'):
+               printf("&lt;");
+               break;
+       default:
+               putchar((unsigned char)c);
+               break;
+       }
+}
+static void
+http_printquery(const struct req *req)
+{
+
+       printf("&expr=");
+       http_print(req->q.expr ? req->q.expr : "");
+       printf("&sec=");
+       http_print(req->q.sec ? req->q.sec : "");
+       printf("&arch=");
+       http_print(req->q.arch ? req->q.arch : "");
 }
-#endif
+
 
 static void
-kval_free(struct kval *p, size_t sz)
+html_printquery(const struct req *req)
 {
-       int              i;
 
-       for (i = 0; i < (int)sz; i++) {
-               free(p[i].key);
-               free(p[i].val);
-       }
-       free(p);
+       printf("&amp;expr=");
+       html_print(req->q.expr ? req->q.expr : "");
+       printf("&amp;sec=");
+       html_print(req->q.sec ? req->q.sec : "");
+       printf("&amp;arch=");
+       html_print(req->q.arch ? req->q.arch : "");
+}
+
+static void
+http_print(const char *p)
+{
+
+       if (NULL == p)
+               return;
+       while ('\0' != *p)
+               http_putchar(*p++);
+}
+
+/*
+ * Call through to html_putchar().
+ * Accepts NULL strings.
+ */
+static void
+html_print(const char *p)
+{
+       
+       if (NULL == p)
+               return;
+       while ('\0' != *p)
+               html_putchar(*p++);
 }
 
 /*
  * Parse out key-value pairs from an HTTP request variable.
- * This can be either a cookie or a POST/GET string.
+ * This can be either a cookie or a POST/GET string, although man.cgi
+ * uses only GET for simplicity.
  */
 static void
-kval_parse(struct kval **kv, size_t *kvsz, char *p)
+http_parse(struct req *req, char *p)
 {
-       char            *key, *val;
-       size_t           sz, cur;
+       char            *key, *val, *manroot;
+       int              i, legacy;
 
-       cur = 0;
+       memset(&req->q, 0, sizeof(struct query));
 
-       while (p && '\0' != *p) {
-               while (' ' == *p)
-                       p++;
+       legacy = -1;
+       manroot = NULL;
 
+       while ('\0' != *p) {
                key = p;
                val = NULL;
 
-               if (NULL != (p = strchr(p, '='))) {
+               p += (int)strcspn(p, ";&");
+               if ('\0' != *p)
                        *p++ = '\0';
-                       val = p;
-
-                       sz = strcspn(p, ";&");
-                       /* LINTED */
-                       p += sz;
-
-                       if ('\0' != *p)
-                               *p++ = '\0';
-               } else {
-                       p = key;
-                       sz = strcspn(p, ";&");
-                       /* LINTED */
-                       p += sz;
-
-                       if ('\0' != *p)
-                               p++;
-                       continue;
-               }
+               if (NULL != (val = strchr(key, '=')))
+                       *val++ = '\0';
 
-               if ('\0' == *key || '\0' == *val)
+               if ('\0' == *key || NULL == val || '\0' == *val)
                        continue;
 
                /* Just abort handling. */
 
-               if ( ! kval_decode(key))
-                       return;
-               if ( ! kval_decode(val))
-                       return;
+               if ( ! http_decode(key))
+                       break;
+               if (NULL != val && ! http_decode(val))
+                       break;
 
-               if (*kvsz + 1 >= cur) {
-                       cur++;
-                       *kv = mandoc_realloc
-                               (*kv, cur * sizeof(struct kval));
-               }
+               if (0 == strcmp(key, "expr"))
+                       req->q.expr = val;
+               else if (0 == strcmp(key, "query"))
+                       req->q.expr = val;
+               else if (0 == strcmp(key, "sec"))
+                       req->q.sec = val;
+               else if (0 == strcmp(key, "sektion"))
+                       req->q.sec = val;
+               else if (0 == strcmp(key, "arch"))
+                       req->q.arch = val;
+               else if (0 == strcmp(key, "manpath"))
+                       manroot = val;
+               else if (0 == strcmp(key, "apropos"))
+                       legacy = 0 == strcmp(val, "0");
+       }
+
+       /* Test for old man.cgi compatibility mode. */
+
+       req->q.legacy = legacy > 0;
+
+       /* 
+        * Section "0" means no section when in legacy mode.
+        * For some man.cgi scripts, "default" arch is none.
+        */
+
+       if (req->q.legacy && NULL != req->q.sec)
+               if (0 == strcmp(req->q.sec, "0"))
+                       req->q.sec = NULL;
+       if (req->q.legacy && NULL != req->q.arch)
+               if (0 == strcmp(req->q.arch, "default"))
+                       req->q.arch = NULL;
+
+       /* Default to first manroot. */
+
+       if (NULL != manroot) {
+               for (i = 0; i < (int)req->psz; i++)
+                       if (0 == strcmp(req->p[i].name, manroot))
+                               break;
+               req->q.manroot = i < (int)req->psz ? i : -1;
+       }
+}
+
+static void
+http_putchar(char c)
+{
 
-               (*kv)[(int)*kvsz].key = mandoc_strdup(key);
-               (*kv)[(int)*kvsz].val = mandoc_strdup(val);
-               (*kvsz)++;
+       if (isalnum((unsigned char)c)) {
+               putchar((unsigned char)c);
+               return;
+       } else if (' ' == c) {
+               putchar('+');
+               return;
        }
+       printf("%%%.2x", c);
 }
 
 /*
- * In-place HTTP-decode a string.  The standard explanation is that this
- * turns "%4e+foo" into "n foo" in the regular way.  This is done
- * in-place over the allocated string.
+ * HTTP-decode a string.  The standard explanation is that this turns
+ * "%4e+foo" into "n foo" in the regular way.  This is done in-place
+ * over the allocated string.
  */
 static int
-kval_decode(char *p)
+http_decode(char *p)
 {
        char             hex[3];
        int              c;
@@ -203,89 +349,711 @@ kval_decode(char *p)
        return(1);
 }
 
+static void
+resp_begin_http(int code, const char *msg)
+{
+
+       if (200 != code)
+               printf("Status: %d %s\n", code, msg);
+
+       puts("Content-Type: text/html; charset=utf-8\n"
+            "Cache-Control: no-cache\n"
+            "Pragma: no-cache\n"
+            "");
+
+       fflush(stdout);
+}
 
-/* ARGSUSED */
 static void
-pg_index(const struct req *req, char *path)
+resp_begin_html(int code, const char *msg)
 {
 
+       resp_begin_http(code, msg);
+
+       printf("<!DOCTYPE HTML PUBLIC "
+              " \"-//W3C//DTD HTML 4.01//EN\""
+              " \"http://www.w3.org/TR/html4/strict.dtd\">\n"
+              "<HTML>\n"
+              "<HEAD>\n"
+              "<META HTTP-EQUIV=\"Content-Type\""
+              " CONTENT=\"text/html; charset=utf-8\">\n"
+              "<LINK REL=\"stylesheet\" HREF=\"%s/man-cgi.css\""
+              " TYPE=\"text/css\" media=\"all\">\n"
+              "<LINK REL=\"stylesheet\" HREF=\"%s/man.css\""
+              " TYPE=\"text/css\" media=\"all\">\n"
+              "<TITLE>System Manpage Reference</TITLE>\n"
+              "</HEAD>\n"
+              "<BODY>\n"
+              "<!-- Begin page content. //-->\n", css, css);
 }
 
-#if 0
 static void
-pg_searchres(struct rec *recs, size_t sz, void *arg)
+resp_end_html(void)
+{
+
+       puts("</BODY>\n"
+            "</HTML>");
+}
+
+static void
+resp_searchform(const struct req *req)
 {
        int              i;
-       const char      *pg;
 
-       if (NULL == (pg = getenv("SCRIPT_NAME")))
-               pg = "";
+       puts("<!-- Begin search form. //-->");
+       printf("<DIV ID=\"mancgi\">\n"
+              "<FORM ACTION=\"%s/search.html\" METHOD=\"get\">\n"
+              "<FIELDSET>\n"
+              "<LEGEND>Search Parameters</LEGEND>\n"
+              "<INPUT TYPE=\"submit\" "
+              " VALUE=\"Search\"> for manuals satisfying \n"
+              "<INPUT TYPE=\"text\" NAME=\"expr\" VALUE=\"",
+              progname);
+       html_print(req->q.expr ? req->q.expr : "");
+       printf("\">, section "
+              "<INPUT TYPE=\"text\""
+              " SIZE=\"4\" NAME=\"sec\" VALUE=\"");
+       html_print(req->q.sec ? req->q.sec : "");
+       printf("\">, arch "
+              "<INPUT TYPE=\"text\""
+              " SIZE=\"8\" NAME=\"arch\" VALUE=\"");
+       html_print(req->q.arch ? req->q.arch : "");
+       printf("\">");
+       if (req->psz > 1) {
+               puts(", <SELECT NAME=\"manpath\">");
+               for (i = 0; i < (int)req->psz; i++) {
+                       printf("<OPTION %s VALUE=\"",
+                               (i == req->q.manroot) ||
+                               (0 == i && -1 == req->q.manroot) ?
+                               "SELECTED=\"selected\"" : "");
+                       html_print(req->p[i].name);
+                       printf("\">");
+                       html_print(req->p[i].name);
+                       puts("</OPTION>");
+               }
+               puts("</SELECT>");
+       }
+       puts(".\n"
+            "<INPUT TYPE=\"reset\" VALUE=\"Reset\">\n"
+            "</FIELDSET>\n"
+            "</FORM>\n"
+            "</DIV>");
+       puts("<!-- End search form. //-->");
+}
+
+static void
+resp_index(const struct req *req)
+{
+
+       resp_begin_html(200, NULL);
+       resp_searchform(req);
+       resp_end_html();
+}
+
+static void
+resp_error400(void)
+{
+
+       resp_begin_html(400, "Query Malformed");
+       printf("<H1>Malformed Query</H1>\n"
+              "<P>\n"
+              "The query your entered was malformed.\n"
+              "Try again from the\n"
+              "<A HREF=\"%s/index.html\">main page</A>.\n"
+              "</P>", progname);
+       resp_end_html();
+}
+
+static void
+resp_error404(const char *page)
+{
 
-       for (i = 0; i < (int)sz; i++) {
-               printf("<A HREF=\"%s/show/%u.html\">", 
-                               pg, recs[i].rec);
-               html_printtext(recs[i].title);
+       resp_begin_html(404, "Not Found");
+       puts("<H1>Page Not Found</H1>\n"
+            "<P>\n"
+            "The page you're looking for, ");
+       printf("<B>");
+       html_print(page);
+       printf("</B>,\n"
+              "could not be found.\n"
+              "Try searching from the\n"
+              "<A HREF=\"%s/index.html\">main page</A>.\n"
+              "</P>", progname);
+       resp_end_html();
+}
+
+static void
+resp_bad(void)
+{
+       resp_begin_html(500, "Internal Server Error");
+       puts("<P>Generic badness happened.</P>");
+       resp_end_html();
+}
+
+static void
+resp_baddb(void)
+{
+
+       resp_begin_html(500, "Internal Server Error");
+       puts("<P>Your database is broken.</P>");
+       resp_end_html();
+}
+
+static void
+resp_search(struct res *r, size_t sz, void *arg)
+{
+       size_t           i, matched;
+       const struct req *req;
+
+       req = (const struct req *)arg;
+
+       if (sz > 0)
+               assert(req->q.manroot >= 0);
+
+       for (matched = i = 0; i < sz; i++)
+               if (r[i].matched)
+                       matched++;
+       
+       if (1 == matched) {
+               for (i = 0; i < sz; i++)
+                       if (r[i].matched)
+                               break;
+               /*
+                * If we have just one result, then jump there now
+                * without any delay.
+                */
+               puts("Status: 303 See Other");
+               printf("Location: http://%s%s/show/%d/%u/%u.html?",
+                               host, progname, req->q.manroot,
+                               r[i].volume, r[i].rec);
+               http_printquery(req);
+               puts("\n"
+                    "Content-Type: text/html; charset=utf-8\n");
+               return;
+       }
+
+       resp_begin_html(200, NULL);
+       resp_searchform(req);
+
+       puts("<DIV CLASS=\"results\">");
+
+       if (0 == matched) {
+               puts("<P>\n"
+                    "No results found.\n"
+                    "</P>\n"
+                    "</DIV>");
+               resp_end_html();
+               return;
+       }
+
+       qsort(r, sz, sizeof(struct res), cmp);
+
+       puts("<TABLE>");
+
+       for (i = 0; i < sz; i++) {
+               if ( ! r[i].matched)
+                       continue;
+               printf("<TR>\n"
+                      "<TD CLASS=\"title\">\n"
+                      "<A HREF=\"%s/show/%d/%u/%u.html?", 
+                               progname, req->q.manroot,
+                               r[i].volume, r[i].rec);
+               html_printquery(req);
+               printf("\">");
+               html_print(r[i].title);
                putchar('(');
-               html_printtext(recs[i].cat);
-               puts(")</A>");
+               html_print(r[i].cat);
+               if (r[i].arch && '\0' != *r[i].arch) {
+                       putchar('/');
+                       html_print(r[i].arch);
+               }
+               printf(")</A>\n"
+                      "</TD>\n"
+                      "<TD CLASS=\"desc\">");
+               html_print(r[i].desc);
+               puts("</TD>\n"
+                    "</TR>");
        }
+
+       puts("</TABLE>\n"
+            "</DIV>");
+       resp_end_html();
 }
-#endif
 
+/* ARGSUSED */
 static void
-pg_search(const struct req *req, char *path)
+pg_index(const struct req *req, char *path)
+{
+
+       resp_index(req);
+}
+
+static void
+catman(const struct req *req, const char *file)
 {
+       FILE            *f;
+       size_t           len;
        int              i;
-       struct opts      opt;
+       char            *p;
+       int              italic, bold;
 
-       for (i = 0; i < (int)req->fieldsz; i++)
-               if (0 == strcmp(req->fields[i].key, "key"))
-                       break;
+       if (NULL == (f = fopen(file, "r"))) {
+               resp_baddb();
+               return;
+       }
 
-       if (i == (int)req->fieldsz)
+       resp_begin_html(200, NULL);
+       resp_searchform(req);
+       puts("<DIV CLASS=\"catman\">\n"
+            "<PRE>");
+
+       while (NULL != (p = fgetln(f, &len))) {
+               bold = italic = 0;
+               for (i = 0; i < (int)len - 1; i++) {
+                       /* 
+                        * This means that the catpage is out of state.
+                        * Ignore it and keep going (although the
+                        * catpage is bogus).
+                        */
+
+                       if ('\b' == p[i] || '\n' == p[i])
+                               continue;
+
+                       /*
+                        * Print a regular character.
+                        * Close out any bold/italic scopes.
+                        * If we're in back-space mode, make sure we'll
+                        * have something to enter when we backspace.
+                        */
+
+                       if ('\b' != p[i + 1]) {
+                               if (italic)
+                                       printf("</I>");
+                               if (bold)
+                                       printf("</B>");
+                               italic = bold = 0;
+                               html_putchar(p[i]);
+                               continue;
+                       } else if (i + 2 >= (int)len)
+                               continue;
+
+                       /* Italic mode. */
+
+                       if ('_' == p[i]) {
+                               if (bold)
+                                       printf("</B>");
+                               if ( ! italic)
+                                       printf("<I>");
+                               bold = 0;
+                               italic = 1;
+                               i += 2;
+                               html_putchar(p[i]);
+                               continue;
+                       }
+
+                       /* 
+                        * Handle funny behaviour troff-isms.
+                        * These grok'd from the original man2html.c.
+                        */
+
+                       if (('+' == p[i] && 'o' == p[i + 2]) ||
+                                       ('o' == p[i] && '+' == p[i + 2]) ||
+                                       ('|' == p[i] && '=' == p[i + 2]) ||
+                                       ('=' == p[i] && '|' == p[i + 2]) ||
+                                       ('*' == p[i] && '=' == p[i + 2]) ||
+                                       ('=' == p[i] && '*' == p[i + 2]) ||
+                                       ('*' == p[i] && '|' == p[i + 2]) ||
+                                       ('|' == p[i] && '*' == p[i + 2]))  {
+                               if (italic)
+                                       printf("</I>");
+                               if (bold)
+                                       printf("</B>");
+                               italic = bold = 0;
+                               putchar('*');
+                               i += 2;
+                               continue;
+                       } else if (('|' == p[i] && '-' == p[i + 2]) ||
+                                       ('-' == p[i] && '|' == p[i + 1]) ||
+                                       ('+' == p[i] && '-' == p[i + 1]) ||
+                                       ('-' == p[i] && '+' == p[i + 1]) ||
+                                       ('+' == p[i] && '|' == p[i + 1]) ||
+                                       ('|' == p[i] && '+' == p[i + 1]))  {
+                               if (italic)
+                                       printf("</I>");
+                               if (bold)
+                                       printf("</B>");
+                               italic = bold = 0;
+                               putchar('+');
+                               i += 2;
+                               continue;
+                       }
+
+                       /* Bold mode. */
+                       
+                       if (italic)
+                               printf("</I>");
+                       if ( ! bold)
+                               printf("<B>");
+                       bold = 1;
+                       italic = 0;
+                       i += 2;
+                       html_putchar(p[i]);
+               }
+
+               /* 
+                * Clean up the last character.
+                * We can get to a newline; don't print that. 
+                */
+
+               if (italic)
+                       printf("</I>");
+               if (bold)
+                       printf("</B>");
+
+               if (i == (int)len - 1 && '\n' != p[i])
+                       html_putchar(p[i]);
+
+               putchar('\n');
+       }
+
+       puts("</PRE>\n"
+            "</DIV>\n"
+            "</BODY>\n"
+            "</HTML>");
+
+       fclose(f);
+}
+
+static void
+format(const struct req *req, const char *file)
+{
+       struct mparse   *mp;
+       int              fd;
+       struct mdoc     *mdoc;
+       struct man      *man;
+       void            *vp;
+       enum mandoclevel rc;
+       char             opts[PATH_MAX + 128];
+
+       if (-1 == (fd = open(file, O_RDONLY, 0))) {
+               resp_baddb();
+               return;
+       }
+
+       mp = mparse_alloc(MPARSE_SO, MANDOCLEVEL_FATAL, NULL, NULL);
+       rc = mparse_readfd(mp, fd, file);
+       close(fd);
+
+       if (rc >= MANDOCLEVEL_FATAL) {
+               resp_baddb();
+               return;
+       }
+
+       snprintf(opts, sizeof(opts), "fragment,"
+                       "man=%s/search.html?sec=%%S&expr=Nm~^%%N$,"
+                       /*"includes=/cgi-bin/man.cgi/usr/include/%%I"*/,
+                       progname);
+
+       mparse_result(mp, &mdoc, &man, NULL);
+       if (NULL == man && NULL == mdoc) {
+               resp_baddb();
+               mparse_free(mp);
                return;
+       }
+
+       resp_begin_html(200, NULL);
+       resp_searchform(req);
+
+       vp = html_alloc(opts);
+
+       if (NULL != mdoc)
+               html_mdoc(vp, mdoc);
+       else
+               html_man(vp, man);
+
+       puts("</BODY>\n"
+            "</HTML>");
+
+       html_free(vp);
+       mparse_free(mp);
+}
+
+static void
+pg_show(const struct req *req, char *path)
+{
+       struct manpaths  ps;
+       size_t           sz;
+       char            *sub;
+       char             file[PATH_MAX];
+       const char      *cp;
+       int              rc, catm;
+       unsigned int     vol, rec, mr;
+       DB              *idx;
+       DBT              key, val;
+
+       idx = NULL;
+
+       /* Parse out mroot, volume, and record from the path. */
+
+       if (NULL == path || NULL == (sub = strchr(path, '/'))) {
+               resp_error400();
+               return;
+       } 
+       *sub++ = '\0';
+       if ( ! atou(path, &mr)) {
+               resp_error400();
+               return;
+       }
+       path = sub;
+       if (NULL == (sub = strchr(path, '/'))) {
+               resp_error400();
+               return;
+       }
+       *sub++ = '\0';
+       if ( ! atou(path, &vol) || ! atou(sub, &rec)) {
+               resp_error400();
+               return;
+       } else if (mr >= (unsigned int)req->psz) {
+               resp_error400();
+               return;
+       }
+
+       /*
+        * Begin by chdir()ing into the manroot.
+        * This way we can pick up the database files, which are
+        * relative to the manpath root.
+        */
+
+       if (-1 == chdir(req->p[(int)mr].path)) {
+               perror(req->p[(int)mr].path);
+               resp_baddb();
+               return;
+       }
+
+       memset(&ps, 0, sizeof(struct manpaths));
+       manpath_manconf(&ps, "etc/catman.conf");
+
+       if (vol >= (unsigned int)ps.sz) {
+               resp_error400();
+               goto out;
+       }
+
+       sz = strlcpy(file, ps.paths[vol], PATH_MAX);
+       assert(sz < PATH_MAX);
+       strlcat(file, "/", PATH_MAX);
+       strlcat(file, MANDOC_IDX, PATH_MAX);
+
+       /* Open the index recno(3) database. */
+
+       idx = dbopen(file, O_RDONLY, 0, DB_RECNO, NULL);
+       if (NULL == idx) {
+               perror(file);
+               resp_baddb();
+               goto out;
+       }
+
+       key.data = &rec;
+       key.size = 4;
+
+       if (0 != (rc = (*idx->get)(idx, &key, &val, 0))) {
+               rc < 0 ? resp_baddb() : resp_error400();
+               goto out;
+       } else if (0 == val.size) {
+               resp_baddb();
+               goto out;
+       }
+
+       cp = (char *)val.data;
+       catm = 'c' == *cp++;
+
+       if (NULL == memchr(cp, '\0', val.size - 1)) 
+               resp_baddb();
+       else {
+               file[(int)sz] = '\0';
+               strlcat(file, "/", PATH_MAX);
+               strlcat(file, cp, PATH_MAX);
+               if (catm) 
+                       catman(req, file);
+               else
+                       format(req, file);
+       }
+out:
+       if (idx)
+               (*idx->close)(idx);
+       manpath_free(&ps);
+}
+
+static void
+pg_search(const struct req *req, char *path)
+{
+       size_t            tt, ressz;
+       struct manpaths   ps;
+       int               i, sz, rc;
+       const char       *ep, *start;
+       struct res      *res;
+       char            **cp;
+       struct opts       opt;
+       struct expr      *expr;
+
+       if (req->q.manroot < 0 || 0 == req->psz) {
+               resp_search(NULL, 0, (void *)req);
+               return;
+       }
 
        memset(&opt, 0, sizeof(struct opts));
-       /*opt.types = TYPE_NAME | TYPE_DESC;
-       apropos_search(&opt, req->fields[i].val, NULL, pg_searchres);*/
+
+       ep       = req->q.expr;
+       opt.arch = req->q.arch;
+       opt.cat  = req->q.sec;
+       rc       = -1;
+       sz       = 0;
+       cp       = NULL;
+       ressz    = 0;
+       res      = NULL;
+
+       /*
+        * Begin by chdir()ing into the root of the manpath.
+        * This way we can pick up the database files, which are
+        * relative to the manpath root.
+        */
+
+       assert(req->q.manroot < (int)req->psz);
+       if (-1 == (chdir(req->p[req->q.manroot].path))) {
+               perror(req->p[req->q.manroot].path);
+               resp_search(NULL, 0, (void *)req);
+               return;
+       }
+
+       memset(&ps, 0, sizeof(struct manpaths));
+       manpath_manconf(&ps, "etc/catman.conf");
+
+       /*
+        * Poor man's tokenisation: just break apart by spaces.
+        * Yes, this is half-ass.  But it works for now.
+        */
+
+       while (ep && isspace((unsigned char)*ep))
+               ep++;
+
+       while (ep && '\0' != *ep) {
+               cp = mandoc_realloc(cp, (sz + 1) * sizeof(char *));
+               start = ep;
+               while ('\0' != *ep && ! isspace((unsigned char)*ep))
+                       ep++;
+               cp[sz] = mandoc_malloc((ep - start) + 1);
+               memcpy(cp[sz], start, ep - start);
+               cp[sz++][ep - start] = '\0';
+               while (isspace((unsigned char)*ep))
+                       ep++;
+       }
+
+       /*
+        * Pump down into apropos backend.
+        * The resp_search() function is called with the results.
+        */
+
+       expr = req->q.legacy ? 
+               termcomp(sz, cp, &tt) : exprcomp(sz, cp, &tt);
+
+       if (NULL != expr)
+               rc = apropos_search
+                       (ps.sz, ps.paths, &opt, expr, tt, 
+                        (void *)req, &ressz, &res, resp_search);
+
+       /* ...unless errors occured. */
+
+       if (0 == rc)
+               resp_baddb();
+       else if (-1 == rc)
+               resp_search(NULL, 0, NULL);
+
+       for (i = 0; i < sz; i++)
+               free(cp[i]);
+
+       free(cp);
+       resfree(res, ressz);
+       exprfree(expr);
+       manpath_free(&ps);
 }
 
 int
 main(void)
 {
        int              i;
+       char             buf[PATH_MAX];
+       DIR             *cwd;
        struct req       req;
-       char            *p;
-       char            *path, *subpath, *suffix;
+       char            *p, *path, *subpath;
+
+       /* Scan our run-time environment. */
+
+       if (NULL == (cache = getenv("CACHE_DIR")))
+               cache = "/cache/man.cgi";
+
+       if (NULL == (progname = getenv("SCRIPT_NAME")))
+               progname = "";
+
+       if (NULL == (css = getenv("CSS_DIR")))
+               css = "";
+
+       if (NULL == (host = getenv("HTTP_HOST")))
+               host = "localhost";
+
+       /*
+        * First we change directory into the cache directory so that
+        * subsequent scanning for manpath directories is rooted
+        * relative to the same position.
+        */
+
+       if (-1 == chdir(cache)) {
+               perror(cache);
+               resp_bad();
+               return(EXIT_FAILURE);
+       } else if (NULL == (cwd = opendir(cache))) {
+               perror(cache);
+               resp_bad();
+               return(EXIT_FAILURE);
+       } 
 
        memset(&req, 0, sizeof(struct req));
 
+       strlcpy(buf, ".", PATH_MAX);
+       pathgen(cwd, buf, &req);
+       closedir(cwd);
+
+       /* Next parse out the query string. */
+
        if (NULL != (p = getenv("QUERY_STRING")))
-               kval_parse(&req.fields, &req.fieldsz, p);
+               http_parse(&req, p);
 
-       suffix = subpath = path = NULL;
+       /*
+        * Now juggle paths to extract information.
+        * We want to extract our filetype (the file suffix), the
+        * initial path component, then the trailing component(s).
+        * Start with leading subpath component. 
+        */
 
-       req.media = MEDIA_HTML;
+       subpath = path = NULL;
        req.page = PAGE__MAX;
 
        if (NULL == (path = getenv("PATH_INFO")) || '\0' == *path)
                req.page = PAGE_INDEX;
+
        if (NULL != path && '/' == *path && '\0' == *++path)
                req.page = PAGE_INDEX;
 
-       if (NULL != path && NULL != (suffix = strrchr(path, '.')))
-               if (NULL != suffix && NULL == strchr(suffix, '/'))
-                       *suffix++ = '\0';
+       /* Strip file suffix. */
+
+       if (NULL != path && NULL != (p = strrchr(path, '.')))
+               if (NULL != p && NULL == strchr(p, '/'))
+                       *p++ = '\0';
+
+       /* Resolve subpath component. */
 
        if (NULL != path && NULL != (subpath = strchr(path, '/')))
-                       *subpath++ = '\0';
+               *subpath++ = '\0';
 
-       if (NULL != suffix && '\0' != *suffix)
-               for (i = 0; i < (int)MEDIA__MAX; i++)
-                       if (0 == strcmp(medias[i], suffix)) {
-                               req.media = (enum media)i;
-                               break;
-                       }
+       /* Map path into one we recognise. */
 
        if (NULL != path && '\0' != *path)
                for (i = 0; i < (int)PAGE__MAX; i++) 
@@ -294,6 +1062,8 @@ main(void)
                                break;
                        }
 
+       /* Route pages. */
+
        switch (req.page) {
        case (PAGE_INDEX):
                pg_index(&req, subpath);
@@ -301,11 +1071,164 @@ main(void)
        case (PAGE_SEARCH):
                pg_search(&req, subpath);
                break;
+       case (PAGE_SHOW):
+               pg_show(&req, subpath);
+               break;
        default:
-               /* Blah */
+               resp_error404(path);
                break;
        }
 
-       kval_free(req.fields, req.fieldsz);
+       for (i = 0; i < (int)req.psz; i++) {
+               free(req.p[i].path);
+               free(req.p[i].name);
+       }
+
+       free(req.p);
        return(EXIT_SUCCESS);
 }
+
+static int
+cmp(const void *p1, const void *p2)
+{
+
+       return(strcasecmp(((const struct res *)p1)->title,
+                               ((const struct res *)p2)->title));
+}
+
+/*
+ * Check to see if an "etc" path consists of a catman.conf file.  If it
+ * does, that means that the path contains a tree created by catman(8)
+ * and should be used for indexing.
+ */
+static int
+pathstop(DIR *dir)
+{
+       struct dirent   *d;
+#if defined(__sun)
+       struct stat      sb;
+#endif
+
+       while (NULL != (d = readdir(dir))) {
+#if defined(__sun)
+               stat(d->d_name, &sb);
+               if (S_IFREG & sb.st_mode)
+#else
+               if (DT_REG == d->d_type)
+#endif
+                       if (0 == strcmp(d->d_name, "catman.conf"))
+                               return(1);
+  }
+
+       return(0);
+}
+
+/*
+ * Scan for indexable paths.
+ * This adds all paths with "etc/catman.conf" to the buffer.
+ */
+static void
+pathgen(DIR *dir, char *path, struct req *req)
+{
+       struct dirent   *d;
+       char            *cp;
+       DIR             *cd;
+       int              rc;
+       size_t           sz, ssz;
+#if defined(__sun)
+       struct stat      sb;
+#endif
+
+       sz = strlcat(path, "/", PATH_MAX);
+       if (sz >= PATH_MAX) {
+               fprintf(stderr, "%s: Path too long", path);
+               return;
+       } 
+
+       /* 
+        * First, scan for the "etc" directory.
+        * If it's found, then see if it should cause us to stop.  This
+        * happens when a catman.conf is found in the directory.
+        */
+
+       rc = 0;
+       while (0 == rc && NULL != (d = readdir(dir))) {
+#if defined(__sun)
+               stat(d->d_name, &sb);
+               if (!(S_IFDIR & sb.st_mode)
+#else
+               if (DT_DIR != d->d_type
+#endif
+        || strcmp(d->d_name, "etc"))
+                       continue;
+
+               path[(int)sz] = '\0';
+               ssz = strlcat(path, d->d_name, PATH_MAX);
+
+               if (ssz >= PATH_MAX) {
+                       fprintf(stderr, "%s: Path too long", path);
+                       return;
+               } else if (NULL == (cd = opendir(path))) {
+                       perror(path);
+                       return;
+               } 
+               
+               rc = pathstop(cd);
+               closedir(cd);
+       }
+
+       if (rc > 0) {
+               /* This also strips the trailing slash. */
+               path[(int)--sz] = '\0';
+               req->p = mandoc_realloc
+                       (req->p, 
+                        (req->psz + 1) * sizeof(struct paths));
+               /*
+                * Strip out the leading "./" unless we're just a ".",
+                * in which case use an empty string as our name.
+                */
+               req->p[(int)req->psz].path = mandoc_strdup(path);
+               req->p[(int)req->psz].name = 
+                       cp = mandoc_strdup(path + (1 == sz ? 1 : 2));
+               req->psz++;
+               /* 
+                * The name is just the path with all the slashes taken
+                * out of it.  Simple but effective. 
+                */
+               for ( ; '\0' != *cp; cp++) 
+                       if ('/' == *cp)
+                               *cp = ' ';
+               return;
+       } 
+
+       /*
+        * If no etc/catman.conf was found, recursively enter child
+        * directory and continue scanning.
+        */
+
+       rewinddir(dir);
+       while (NULL != (d = readdir(dir))) {
+#if defined(__sun)
+               stat(d->d_name, &sb);
+               if (!(S_IFDIR & sb.st_mode)
+#else
+               if (DT_DIR != d->d_type
+#endif
+        || '.' == d->d_name[0])
+                       continue;
+
+               path[(int)sz] = '\0';
+               ssz = strlcat(path, d->d_name, PATH_MAX);
+
+               if (ssz >= PATH_MAX) {
+                       fprintf(stderr, "%s: Path too long", path);
+                       return;
+               } else if (NULL == (cd = opendir(path))) {
+                       perror(path);
+                       return;
+               }
+
+               pathgen(cd, path, req);
+               closedir(cd);
+       }
+}