]> git.cameronkatri.com Git - mandoc.git/blobdiff - main.c
Merge OpenBSD's `so' handling (plus some documentation). Great work to
[mandoc.git] / main.c
diff --git a/main.c b/main.c
index c672f621a0c3ed599611343b00e151701ddbb6e3..16b54def251c26cfa738942ef75d5e06b3d1127f 100644 (file)
--- a/main.c
+++ b/main.c
@@ -1,4 +1,4 @@
-/*     $Id: main.c,v 1.97 2010/07/04 22:04:04 schwarze Exp $ */
+/*     $Id: main.c,v 1.113 2010/12/01 16:54:25 kristaps Exp $ */
 /*
  * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
  * Copyright (c) 2010 Ingo Schwarze <schwarze@openbsd.org>
 /*
  * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
  * Copyright (c) 2010 Ingo Schwarze <schwarze@openbsd.org>
@@ -23,6 +23,7 @@
 #include <sys/stat.h>
 
 #include <assert.h>
 #include <sys/stat.h>
 
 #include <assert.h>
+#include <ctype.h>
 #include <fcntl.h>
 #include <stdio.h>
 #include <stdint.h>
 #include <fcntl.h>
 #include <stdio.h>
 #include <stdint.h>
 #include <unistd.h>
 
 #include "mandoc.h"
 #include <unistd.h>
 
 #include "mandoc.h"
-#include "regs.h"
 #include "main.h"
 #include "mdoc.h"
 #include "man.h"
 #include "roff.h"
 
 #include "main.h"
 #include "mdoc.h"
 #include "man.h"
 #include "roff.h"
 
+#ifndef MAP_FILE
+#define        MAP_FILE        0
+#endif
+
 #define        UNCONST(a)      ((void *)(uintptr_t)(const void *)(a))
 
 /* FIXME: Intel's compiler?  LLVM?  pcc?  */
 #define        UNCONST(a)      ((void *)(uintptr_t)(const void *)(a))
 
 /* FIXME: Intel's compiler?  LLVM?  pcc?  */
@@ -68,24 +72,19 @@ enum        outt {
        OUTT_HTML,
        OUTT_XHTML,
        OUTT_LINT,
        OUTT_HTML,
        OUTT_XHTML,
        OUTT_LINT,
-       OUTT_PS
+       OUTT_PS,
+       OUTT_PDF
 };
 
 struct curparse {
        const char       *file;         /* Current parse. */
        int               fd;           /* Current parse. */
 };
 
 struct curparse {
        const char       *file;         /* Current parse. */
        int               fd;           /* Current parse. */
-       int               wflags;
-       /* FIXME: set by max error */
-#define        WARN_WALL        (1 << 0)       /* All-warnings mask. */
-#define        WARN_WERR        (1 << 2)       /* Warnings->errors. */
-       int               fflags;
-#define        FL_IGN_SCOPE     (1 << 0)       /* Ignore scope errors. */
-#define        FL_NIGN_ESCAPE   (1 << 1)       /* Don't ignore bad escapes. */
-#define        FL_NIGN_MACRO    (1 << 2)       /* Don't ignore bad macros. */
-#define        FL_IGN_ERRORS    (1 << 4)       /* Ignore failed parse. */
-#define        FL_STRICT         FL_NIGN_ESCAPE | \
-                         FL_NIGN_MACRO /* ignore nothing */
+       int               line;         /* Line number in the file. */
+       enum mandoclevel  wlevel;       /* Ignore messages below this. */
+       int               wstop;        /* Stop after a file with a warning. */
        enum intt         inttype;      /* which parser to use */
        enum intt         inttype;      /* which parser to use */
+       struct man       *pman;         /* persistent man parser */
+       struct mdoc      *pmdoc;        /* persistent mdoc parser */
        struct man       *man;          /* man parser */
        struct mdoc      *mdoc;         /* mdoc parser */
        struct roff      *roff;         /* roff parser (!NULL) */
        struct man       *man;          /* man parser */
        struct mdoc      *mdoc;         /* mdoc parser */
        struct roff      *roff;         /* roff parser (!NULL) */
@@ -98,31 +97,49 @@ struct      curparse {
        char              outopts[BUFSIZ]; /* buf of output opts */
 };
 
        char              outopts[BUFSIZ]; /* buf of output opts */
 };
 
+static const char * const      mandoclevels[MANDOCLEVEL_MAX] = {
+       "SUCCESS",
+       "RESERVED",
+       "WARNING",
+       "ERROR",
+       "FATAL",
+       "BADARG",
+       "SYSERR"
+};
+
+static const enum mandocerr    mandoclimits[MANDOCLEVEL_MAX] = {
+       MANDOCERR_OK,
+       MANDOCERR_WARNING,
+       MANDOCERR_WARNING,
+       MANDOCERR_ERROR,
+       MANDOCERR_FATAL,
+       MANDOCERR_MAX,
+       MANDOCERR_MAX
+};
+
 static const char * const      mandocerrs[MANDOCERR_MAX] = {
        "ok",
 
        "generic warning",
 
 static const char * const      mandocerrs[MANDOCERR_MAX] = {
        "ok",
 
        "generic warning",
 
+       ".so is fragile, better use ln(1)",
        "text should be uppercase",
        "sections out of conventional order",
        "section name repeats",
        "out of order prologue",
        "repeated prologue entry",
        "list type must come first",
        "text should be uppercase",
        "sections out of conventional order",
        "section name repeats",
        "out of order prologue",
        "repeated prologue entry",
        "list type must come first",
-       "bad standard",
-       "bad library",
+       "tab in non-literal context",
        "bad escape sequence",
        "unterminated quoted string",
        "argument requires the width argument",
        "superfluous width argument",
        "bad escape sequence",
        "unterminated quoted string",
        "argument requires the width argument",
        "superfluous width argument",
-       "ignoring argument",
        "bad date argument",
        "bad width argument",
        "unknown manual section",
        "section not in conventional manual section",
        "end of line whitespace",
        "blocks badly nested",
        "bad date argument",
        "bad width argument",
        "unknown manual section",
        "section not in conventional manual section",
        "end of line whitespace",
        "blocks badly nested",
-       "scope open on exit",
 
        "generic error",
 
 
        "generic error",
 
@@ -130,9 +147,11 @@ static     const char * const      mandocerrs[MANDOCERR_MAX] = {
        "bad Boolean value",
        "child violates parent syntax",
        "bad AT&T symbol",
        "bad Boolean value",
        "child violates parent syntax",
        "bad AT&T symbol",
+       "bad standard",
        "list type repeated",
        "display type repeated",
        "argument repeated",
        "list type repeated",
        "display type repeated",
        "argument repeated",
+       "ignoring argument",
        "manual name not yet set",
        "obsolete macro ignored",
        "empty macro ignored",
        "manual name not yet set",
        "obsolete macro ignored",
        "empty macro ignored",
@@ -144,10 +163,12 @@ static    const char * const      mandocerrs[MANDOCERR_MAX] = {
        "no text in this context",
        "bad comment style",
        "unknown macro will be lost",
        "no text in this context",
        "bad comment style",
        "unknown macro will be lost",
+       "NOT IMPLEMENTED: skipping request",
        "line scope broken",
        "argument count wrong",
        "request scope close w/none open",
        "scope already open",
        "line scope broken",
        "argument count wrong",
        "request scope close w/none open",
        "scope already open",
+       "scope open on exit",
        "macro requires line argument(s)",
        "macro requires body argument(s)",
        "macro requires argument(s)",
        "macro requires line argument(s)",
        "macro requires body argument(s)",
        "macro requires argument(s)",
@@ -157,6 +178,7 @@ static      const char * const      mandocerrs[MANDOCERR_MAX] = {
        "missing font type",
        "line argument(s) will be lost",
        "body argument(s) will be lost",
        "missing font type",
        "line argument(s) will be lost",
        "body argument(s) will be lost",
+       "paragraph macro ignored",
 
        "generic fatal error",
 
 
        "generic fatal error",
 
@@ -165,36 +187,33 @@ static    const char * const      mandocerrs[MANDOCERR_MAX] = {
        "unsupported display type",
        "blocks badly nested",
        "no such block is open",
        "unsupported display type",
        "blocks badly nested",
        "no such block is open",
-       "scope broken, syntax violated",
        "line scope broken, syntax violated",
        "argument count wrong, violates syntax",
        "child violates parent syntax",
        "argument count wrong, violates syntax",
        "line scope broken, syntax violated",
        "argument count wrong, violates syntax",
        "child violates parent syntax",
        "argument count wrong, violates syntax",
+       "NOT IMPLEMENTED: .so with absolute path or \"..\"",
        "no document body",
        "no document prologue",
        "utsname system call failed",
        "no document body",
        "no document prologue",
        "utsname system call failed",
-       "memory exhausted",
+       "static buffer exhausted",
 };
 
 };
 
+static void              parsebuf(struct curparse *, struct buf, int);
+static void              pdesc(struct curparse *);
 static void              fdesc(struct curparse *);
 static void              ffile(const char *, struct curparse *);
 static void              fdesc(struct curparse *);
 static void              ffile(const char *, struct curparse *);
-static int               foptions(int *, char *);
-static struct man       *man_init(struct curparse *);
-static struct mdoc      *mdoc_init(struct curparse *);
-static struct roff      *roff_init(struct curparse *);
+static int               pfile(const char *, struct curparse *);
 static int               moptions(enum intt *, char *);
 static int               mmsg(enum mandocerr, void *, 
                                int, int, const char *);
 static int               moptions(enum intt *, char *);
 static int               mmsg(enum mandocerr, void *, 
                                int, int, const char *);
-static int               pset(const char *, int, struct curparse *,
-                               struct man **, struct mdoc **);
+static void              pset(const char *, int, struct curparse *);
 static int               toptions(struct curparse *, char *);
 static void              usage(void) __attribute__((noreturn));
 static void              version(void) __attribute__((noreturn));
 static int               toptions(struct curparse *, char *);
 static void              usage(void) __attribute__((noreturn));
 static void              version(void) __attribute__((noreturn));
-static int               woptions(int *, char *);
+static int               woptions(struct curparse *, char *);
 
 static const char       *progname;
 
 static const char       *progname;
-static int               with_fatal;
-static int               with_error;
+static enum mandoclevel  exit_status = MANDOCLEVEL_OK;
 
 int
 main(int argc, char *argv[])
 
 int
 main(int argc, char *argv[])
@@ -212,17 +231,14 @@ main(int argc, char *argv[])
 
        curp.inttype = INTT_AUTO;
        curp.outtype = OUTT_ASCII;
 
        curp.inttype = INTT_AUTO;
        curp.outtype = OUTT_ASCII;
+       curp.wlevel  = MANDOCLEVEL_FATAL;
 
        /* LINTED */
 
        /* LINTED */
-       while (-1 != (c = getopt(argc, argv, "f:m:O:T:VW:")))
+       while (-1 != (c = getopt(argc, argv, "m:O:T:VW:")))
                switch (c) {
                switch (c) {
-               case ('f'):
-                       if ( ! foptions(&curp.fflags, optarg))
-                               return(EXIT_FAILURE);
-                       break;
                case ('m'):
                        if ( ! moptions(&curp.inttype, optarg))
                case ('m'):
                        if ( ! moptions(&curp.inttype, optarg))
-                               return(EXIT_FAILURE);
+                               return((int)MANDOCLEVEL_BADARG);
                        break;
                case ('O'):
                        (void)strlcat(curp.outopts, optarg, BUFSIZ);
                        break;
                case ('O'):
                        (void)strlcat(curp.outopts, optarg, BUFSIZ);
@@ -230,11 +246,11 @@ main(int argc, char *argv[])
                        break;
                case ('T'):
                        if ( ! toptions(&curp, optarg))
                        break;
                case ('T'):
                        if ( ! toptions(&curp, optarg))
-                               return(EXIT_FAILURE);
+                               return((int)MANDOCLEVEL_BADARG);
                        break;
                case ('W'):
                        break;
                case ('W'):
-                       if ( ! woptions(&curp.wflags, optarg))
-                               return(EXIT_FAILURE);
+                       if ( ! woptions(&curp, optarg))
+                               return((int)MANDOCLEVEL_BADARG);
                        break;
                case ('V'):
                        version();
                        break;
                case ('V'):
                        version();
@@ -256,23 +272,21 @@ main(int argc, char *argv[])
 
        while (*argv) {
                ffile(*argv, &curp);
 
        while (*argv) {
                ffile(*argv, &curp);
-
-               if (with_fatal && !(curp.fflags & FL_IGN_ERRORS))
+               if (MANDOCLEVEL_OK != exit_status && curp.wstop)
                        break;
                ++argv;
        }
 
        if (curp.outfree)
                (*curp.outfree)(curp.outdata);
                        break;
                ++argv;
        }
 
        if (curp.outfree)
                (*curp.outfree)(curp.outdata);
-       if (curp.mdoc)
-               mdoc_free(curp.mdoc);
-       if (curp.man)
-               man_free(curp.man);
+       if (curp.pmdoc)
+               mdoc_free(curp.pmdoc);
+       if (curp.pman)
+               man_free(curp.pman);
        if (curp.roff)
                roff_free(curp.roff);
 
        if (curp.roff)
                roff_free(curp.roff);
 
-       return((with_fatal || with_error) ? 
-                       EXIT_FAILURE :  EXIT_SUCCESS);
+       return((int)exit_status);
 }
 
 
 }
 
 
@@ -281,7 +295,7 @@ version(void)
 {
 
        (void)printf("%s %s\n", progname, VERSION);
 {
 
        (void)printf("%s %s\n", progname, VERSION);
-       exit(EXIT_SUCCESS);
+       exit((int)MANDOCLEVEL_OK);
 }
 
 
 }
 
 
@@ -289,95 +303,84 @@ static void
 usage(void)
 {
 
 usage(void)
 {
 
-       (void)fprintf(stderr, "usage: %s [-V] [-foption] "
-                       "[-mformat] [-Ooption] [-Toutput] "
-                       "[-Werr] [file...]\n", progname);
-       exit(EXIT_FAILURE);
+       (void)fprintf(stderr, "usage: %s "
+                       "[-V] "
+                       "[-foption] "
+                       "[-mformat] "
+                       "[-Ooption] "
+                       "[-Toutput] "
+                       "[-Werr] "
+                       "[file...]\n", 
+                       progname);
+
+       exit((int)MANDOCLEVEL_BADARG);
 }
 
 }
 
-
-static struct man *
-man_init(struct curparse *curp)
+static void
+ffile(const char *file, struct curparse *curp)
 {
 {
-       int              pflags;
-
-       /* Defaults from mandoc.1. */
-
-       pflags = MAN_IGN_MACRO | MAN_IGN_ESCAPE;
 
 
-       if (curp->fflags & FL_NIGN_MACRO)
-               pflags &= ~MAN_IGN_MACRO;
-       if (curp->fflags & FL_NIGN_ESCAPE)
-               pflags &= ~MAN_IGN_ESCAPE;
+       /*
+        * Called once per input file.  Get the file ready for reading,
+        * pass it through to the parser-driver, then close it out.
+        * XXX: don't do anything special as this is only called for
+        * files; stdin goes directly to fdesc().
+        */
 
 
-       return(man_alloc(&curp->regs, curp, pflags, mmsg));
-}
+       curp->file = file;
 
 
+       if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) {
+               perror(curp->file);
+               exit_status = MANDOCLEVEL_SYSERR;
+               return;
+       }
 
 
-static struct roff *
-roff_init(struct curparse *curp)
-{
+       fdesc(curp);
 
 
-       return(roff_alloc(&curp->regs, mmsg, curp));
+       if (-1 == close(curp->fd))
+               perror(curp->file);
 }
 
 }
 
-
-static struct mdoc *
-mdoc_init(struct curparse *curp)
+static int
+pfile(const char *file, struct curparse *curp)
 {
 {
-       int              pflags;
-
-       /* Defaults from mandoc.1. */
+       const char      *savefile;
+       int              fd, savefd;
 
 
-       pflags = MDOC_IGN_MACRO | MDOC_IGN_ESCAPE;
-
-       if (curp->fflags & FL_IGN_SCOPE)
-               pflags |= MDOC_IGN_SCOPE;
-       if (curp->fflags & FL_NIGN_ESCAPE)
-               pflags &= ~MDOC_IGN_ESCAPE;
-       if (curp->fflags & FL_NIGN_MACRO)
-               pflags &= ~MDOC_IGN_MACRO;
+       if (-1 == (fd = open(file, O_RDONLY, 0))) {
+               perror(file);
+               exit_status = MANDOCLEVEL_SYSERR;
+               return(0);
+       }
 
 
-       return(mdoc_alloc(&curp->regs, curp, pflags, mmsg));
-}
+       savefile = curp->file;
+       savefd = curp->fd;
 
 
+       curp->file = file;
+       curp->fd = fd;
 
 
-static void
-ffile(const char *file, struct curparse *curp)
-{
+       pdesc(curp);
 
 
-       curp->file = file;
-       if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) {
-               perror(curp->file);
-               with_fatal = 1;
-               return;
-       }
+       curp->file = savefile;
+       curp->fd = savefd;
 
 
-       fdesc(curp);
+       if (-1 == close(fd))
+               perror(file);
 
 
-       if (-1 == close(curp->fd))
-               perror(curp->file);
+       return(MANDOCLEVEL_FATAL > exit_status ? 1 : 0);
 }
 
 
 }
 
 
-static int
+static void
 resize_buf(struct buf *buf, size_t initial)
 {
 resize_buf(struct buf *buf, size_t initial)
 {
-       void *tmp;
-       size_t sz;
 
 
-       if (buf->sz == 0)
-               sz = initial;
-       else
-               sz = 2 * buf->sz;
-       tmp = realloc(buf->buf, sz);
-       if (NULL == tmp) {
+       buf->sz = buf->sz ? 2 * buf->sz : initial;
+       buf->buf = realloc(buf->buf, buf->sz);
+       if (NULL == buf->buf) {
                perror(NULL);
                perror(NULL);
-               return(0);
+               exit((int)MANDOCLEVEL_SYSERR);
        }
        }
-       buf->buf = tmp;
-       buf->sz = sz;
-       return(1);
 }
 
 
 }
 
 
@@ -390,7 +393,6 @@ read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap)
 
        if (-1 == fstat(curp->fd, &st)) {
                perror(curp->file);
 
        if (-1 == fstat(curp->fd, &st)) {
                perror(curp->file);
-               with_fatal = 1;
                return(0);
        }
 
                return(0);
        }
 
@@ -405,7 +407,6 @@ read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap)
                if (st.st_size >= (1U << 31)) {
                        fprintf(stderr, "%s: input too large\n", 
                                        curp->file);
                if (st.st_size >= (1U << 31)) {
                        fprintf(stderr, "%s: input too large\n", 
                                        curp->file);
-                       with_fatal = 1;
                        return(0);
                }
                *with_mmap = 1;
                        return(0);
                }
                *with_mmap = 1;
@@ -432,8 +433,7 @@ read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap)
                                                curp->file);
                                break;
                        }
                                                curp->file);
                                break;
                        }
-                       if (! resize_buf(fb, 65536))
-                               break;
+                       resize_buf(fb, 65536);
                }
                ssz = read(curp->fd, fb->buf + (int)off, fb->sz - off);
                if (ssz == 0) {
                }
                ssz = read(curp->fd, fb->buf + (int)off, fb->sz - off);
                if (ssz == 0) {
@@ -449,7 +449,6 @@ read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap)
 
        free(fb->buf);
        fb->buf = NULL;
 
        free(fb->buf);
        fb->buf = NULL;
-       with_fatal = 1;
        return(0);
 }
 
        return(0);
 }
 
@@ -457,56 +456,246 @@ read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap)
 static void
 fdesc(struct curparse *curp)
 {
 static void
 fdesc(struct curparse *curp)
 {
-       struct buf       ln, blk;
-       int              i, pos, lnn, lnn_start, with_mmap, of;
-       enum rofferr     re;
-       struct man      *man;
-       struct mdoc     *mdoc;
-       struct roff     *roff;
 
 
-       man = NULL;
-       mdoc = NULL;
-       roff = NULL;
+       /*
+        * Called once per file with an opened file descriptor.  All
+        * pre-file-parse operations (whether stdin or a file) should go
+        * here.
+        *
+        * This calls down into the nested parser, which drills down and
+        * fully parses a file and all its dependences (i.e., `so').  It
+        * then runs the cleanup validators and pushes to output.
+        */
+
+       /* Zero the parse type. */
 
 
-       memset(&ln, 0, sizeof(struct buf));
+       curp->mdoc = NULL;
+       curp->man = NULL;
+
+       /* Make sure the mandotory roff parser is initialised. */
+
+       if (NULL == curp->roff) {
+               curp->roff = roff_alloc(&curp->regs, curp, mmsg);
+               assert(curp->roff);
+       }
+
+       /* Fully parse the file. */
+
+       pdesc(curp);
+
+       if (MANDOCLEVEL_FATAL <= exit_status)
+               goto cleanup;
+
+       /* NOTE a parser may not have been assigned, yet. */
+
+       if ( ! (curp->man || curp->mdoc)) {
+               fprintf(stderr, "%s: Not a manual\n", curp->file);
+               exit_status = MANDOCLEVEL_FATAL;
+               goto cleanup;
+       }
+
+       /* Clean up the parse routine ASTs. */
+
+       if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
+               assert(MANDOCLEVEL_FATAL <= exit_status);
+               goto cleanup;
+       }
+
+       if (curp->man && ! man_endparse(curp->man)) {
+               assert(MANDOCLEVEL_FATAL <= exit_status);
+               goto cleanup;
+       }
+
+       assert(curp->roff);
+       if ( ! roff_endparse(curp->roff)) {
+               assert(MANDOCLEVEL_FATAL <= exit_status);
+               goto cleanup;
+       }
+
+       /*
+        * With -Wstop and warnings or errors of at least
+        * the requested level, do not produce output.
+        */
+
+       if (MANDOCLEVEL_OK != exit_status && curp->wstop)
+               goto cleanup;
+
+       /* If unset, allocate output dev now (if applicable). */
+
+       if ( ! (curp->outman && curp->outmdoc)) {
+               switch (curp->outtype) {
+               case (OUTT_XHTML):
+                       curp->outdata = xhtml_alloc(curp->outopts);
+                       break;
+               case (OUTT_HTML):
+                       curp->outdata = html_alloc(curp->outopts);
+                       break;
+               case (OUTT_ASCII):
+                       curp->outdata = ascii_alloc(curp->outopts);
+                       curp->outfree = ascii_free;
+                       break;
+               case (OUTT_PDF):
+                       curp->outdata = pdf_alloc(curp->outopts);
+                       curp->outfree = pspdf_free;
+                       break;
+               case (OUTT_PS):
+                       curp->outdata = ps_alloc(curp->outopts);
+                       curp->outfree = pspdf_free;
+                       break;
+               default:
+                       break;
+               }
+
+               switch (curp->outtype) {
+               case (OUTT_HTML):
+                       /* FALLTHROUGH */
+               case (OUTT_XHTML):
+                       curp->outman = html_man;
+                       curp->outmdoc = html_mdoc;
+                       curp->outfree = html_free;
+                       break;
+               case (OUTT_TREE):
+                       curp->outman = tree_man;
+                       curp->outmdoc = tree_mdoc;
+                       break;
+               case (OUTT_PDF):
+                       /* FALLTHROUGH */
+               case (OUTT_ASCII):
+                       /* FALLTHROUGH */
+               case (OUTT_PS):
+                       curp->outman = terminal_man;
+                       curp->outmdoc = terminal_mdoc;
+                       break;
+               default:
+                       break;
+               }
+       }
+
+       /* Execute the out device, if it exists. */
+
+       if (curp->man && curp->outman)
+               (*curp->outman)(curp->outdata, curp->man);
+       if (curp->mdoc && curp->outmdoc)
+               (*curp->outmdoc)(curp->outdata, curp->mdoc);
+
+ cleanup:
+
+       memset(&curp->regs, 0, sizeof(struct regset));
+
+       /* Reset the current-parse compilers. */
+
+       if (curp->mdoc)
+               mdoc_reset(curp->mdoc);
+       if (curp->man)
+               man_reset(curp->man);
+
+       assert(curp->roff);
+       roff_reset(curp->roff);
+
+       return;
+}
+
+static void
+pdesc(struct curparse *curp)
+{
+       struct buf       blk;
+       int              with_mmap;
 
        /*
 
        /*
-        * Two buffers: ln and buf.  buf is the input file and may be
-        * memory mapped.  ln is a line buffer and grows on-demand.
+        * Run for each opened file; may be called more than once for
+        * each full parse sequence if the opened file is nested (i.e.,
+        * from `so').  Simply sucks in the whole file and moves into
+        * the parse phase for the file.
         */
 
         */
 
-       if ( ! read_whole_file(curp, &blk, &with_mmap))
+       if ( ! read_whole_file(curp, &blk, &with_mmap)) {
+               exit_status = MANDOCLEVEL_SYSERR;
                return;
                return;
+       }
 
 
-       if (NULL == curp->roff) 
-               curp->roff = roff_init(curp);
-       if (NULL == (roff = curp->roff))
-               goto bailout;
+       /* Line number is per-file. */
 
 
-       for (i = 0, lnn = 1; i < (int)blk.sz;) {
-               pos = 0;
-               lnn_start = lnn;
-               while (i < (int)blk.sz) {
+       curp->line = 1;
+
+       parsebuf(curp, blk, 1);
+
+       if (with_mmap)
+               munmap(blk.buf, blk.sz);
+       else
+               free(blk.buf);
+}
+
+static void
+parsebuf(struct curparse *curp, struct buf blk, int start)
+{
+       struct buf       ln;
+       int              i, of, rc;
+       int              pos; /* byte number in the ln buffer */
+       int              lnn; /* line number in the real file */
+       unsigned char    c;
+
+       /*
+        * Main parse routine for an opened file.  This is called for
+        * each opened file and simply loops around the full input file,
+        * possibly nesting (i.e., with `so').
+        */
+
+       memset(&ln, 0, sizeof(struct buf));
+
+       lnn = curp->line; 
+       pos = 0; 
+
+       for (i = 0; i < (int)blk.sz; ) {
+               if (0 == pos && '\0' == blk.buf[i])
+                       break;
+
+               if (start)
+                       curp->line = lnn;
+
+               while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
                        if ('\n' == blk.buf[i]) {
                                ++i;
                                ++lnn;
                                break;
                        }
                        if ('\n' == blk.buf[i]) {
                                ++i;
                                ++lnn;
                                break;
                        }
-                       /* Trailing backslash is like a plain character. */
+
+                       /* 
+                        * Warn about bogus characters.  If you're using
+                        * non-ASCII encoding, you're screwing your
+                        * readers.  Since I'd rather this not happen,
+                        * I'll be helpful and drop these characters so
+                        * we don't display gibberish.  Note to manual
+                        * writers: use special characters.
+                        */
+
+                       c = (unsigned char) blk.buf[i];
+
+                       if ( ! (isascii(c) && 
+                                       (isgraph(c) || isblank(c)))) {
+                               mmsg(MANDOCERR_BADCHAR, curp, 
+                                   curp->line, pos, "ignoring byte");
+                               i++;
+                               continue;
+                       }
+
+                       /* Trailing backslash = a plain char. */
+
                        if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
                                if (pos >= (int)ln.sz)
                        if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
                                if (pos >= (int)ln.sz)
-                                       if (! resize_buf(&ln, 256))
-                                               goto bailout;
+                                       resize_buf(&ln, 256);
                                ln.buf[pos++] = blk.buf[i++];
                                continue;
                        }
                                ln.buf[pos++] = blk.buf[i++];
                                continue;
                        }
-                       /* Found an escape and at least one other character. */
+
+                       /* Found escape & at least one other char. */
+
                        if ('\n' == blk.buf[i + 1]) {
                        if ('\n' == blk.buf[i + 1]) {
-                               /* Escaped newlines are skipped over */
                                i += 2;
                                i += 2;
+                               /* Escaped newlines are skipped over */
                                ++lnn;
                                continue;
                        }
                                ++lnn;
                                continue;
                        }
+
                        if ('"' == blk.buf[i + 1]) {
                                i += 2;
                                /* Comment, skip to end of line */
                        if ('"' == blk.buf[i + 1]) {
                                i += 2;
                                /* Comment, skip to end of line */
@@ -517,6 +706,7 @@ fdesc(struct curparse *curp)
                                                break;
                                        }
                                }
                                                break;
                                        }
                                }
+
                                /* Backout trailing whitespaces */
                                for (; pos > 0; --pos) {
                                        if (ln.buf[pos - 1] != ' ')
                                /* Backout trailing whitespaces */
                                for (; pos > 0; --pos) {
                                        if (ln.buf[pos - 1] != ' ')
@@ -526,18 +716,19 @@ fdesc(struct curparse *curp)
                                }
                                break;
                        }
                                }
                                break;
                        }
-                       /* Some other escape sequence, copy and continue. */
+
+                       /* Some other escape sequence, copy & cont. */
+
                        if (pos + 1 >= (int)ln.sz)
                        if (pos + 1 >= (int)ln.sz)
-                               if (! resize_buf(&ln, 256))
-                                       goto bailout;
+                               resize_buf(&ln, 256);
 
                        ln.buf[pos++] = blk.buf[i++];
                        ln.buf[pos++] = blk.buf[i++];
                }
 
                if (pos >= (int)ln.sz)
 
                        ln.buf[pos++] = blk.buf[i++];
                        ln.buf[pos++] = blk.buf[i++];
                }
 
                if (pos >= (int)ln.sz)
-                       if (! resize_buf(&ln, 256))
-                               goto bailout;
+                       resize_buf(&ln, 256);
+
                ln.buf[pos] = '\0';
 
                /*
                ln.buf[pos] = '\0';
 
                /*
@@ -550,15 +741,37 @@ fdesc(struct curparse *curp)
                 */
 
                of = 0;
                 */
 
                of = 0;
-               do {
-                       re = roff_parseln(roff, lnn_start, 
-                                       &ln.buf, &ln.sz, of, &of);
-               } while (ROFF_RERUN == re);
 
 
-               if (ROFF_IGN == re)
+rerun:
+               rc = roff_parseln
+                       (curp->roff, curp->line, 
+                        &ln.buf, &ln.sz, of, &of);
+
+               switch (rc) {
+               case (ROFF_REPARSE):
+                       parsebuf(curp, ln, 0);
+                       pos = 0;
+                       continue;
+               case (ROFF_APPEND):
+                       pos = strlen(ln.buf);
+                       continue;
+               case (ROFF_RERUN):
+                       goto rerun;
+               case (ROFF_IGN):
+                       pos = 0;
                        continue;
                        continue;
-               else if (ROFF_ERR == re)
-                       goto bailout;
+               case (ROFF_ERR):
+                       assert(MANDOCLEVEL_FATAL <= exit_status);
+                       break;
+               case (ROFF_SO):
+                       if (pfile(ln.buf + of, curp)) {
+                               pos = 0;
+                               continue;
+                       } else
+                               break;
+               case (ROFF_CONT):
+                       break;
+               }
 
                /*
                 * If input parsers have not been allocated, do so now.
 
                /*
                 * If input parsers have not been allocated, do so now.
@@ -567,112 +780,43 @@ fdesc(struct curparse *curp)
                 * parsers with each one.
                 */
 
                 * parsers with each one.
                 */
 
-               if ( ! (man || mdoc))
-                       if ( ! pset(ln.buf + of, pos - of, curp, &man, &mdoc))
-                               goto bailout;
-
-               /* Lastly, push down into the parsers themselves. */
+               if ( ! (curp->man || curp->mdoc))
+                       pset(ln.buf + of, pos - of, curp);
 
 
-               if (man && ! man_parseln(man, lnn_start, ln.buf, of))
-                       goto bailout;
-               if (mdoc && ! mdoc_parseln(mdoc, lnn_start, ln.buf, of))
-                       goto bailout;
-       }
-
-       /* NOTE a parser may not have been assigned, yet. */
-
-       if ( ! (man || mdoc)) {
-               fprintf(stderr, "%s: Not a manual\n", curp->file);
-               goto bailout;
-       }
-
-       /* Clean up the parse routine ASTs. */
-
-       if (mdoc && ! mdoc_endparse(mdoc))
-               goto bailout;
-       if (man && ! man_endparse(man))
-               goto bailout;
-       if (roff && ! roff_endparse(roff))
-               goto bailout;
-
-       /* If unset, allocate output dev now (if applicable). */
+               /* 
+                * Lastly, push down into the parsers themselves.  One
+                * of these will have already been set in the pset()
+                * routine.
+                */
 
 
-       if ( ! (curp->outman && curp->outmdoc)) {
-               switch (curp->outtype) {
-               case (OUTT_XHTML):
-                       curp->outdata = xhtml_alloc(curp->outopts);
-                       break;
-               case (OUTT_HTML):
-                       curp->outdata = html_alloc(curp->outopts);
-                       break;
-               case (OUTT_ASCII):
-                       curp->outdata = ascii_alloc(curp->outopts);
-                       curp->outfree = ascii_free;
-                       break;
-               case (OUTT_PS):
-                       curp->outdata = ps_alloc(curp->outopts);
-                       curp->outfree = ps_free;
-                       break;
-               default:
-                       break;
-               }
+               if (curp->man || curp->mdoc) {
+                       rc = curp->man ?
+                               man_parseln(curp->man, 
+                                       curp->line, ln.buf, of) :
+                               mdoc_parseln(curp->mdoc, 
+                                       curp->line, ln.buf, of);
 
 
-               switch (curp->outtype) {
-               case (OUTT_HTML):
-                       /* FALLTHROUGH */
-               case (OUTT_XHTML):
-                       curp->outman = html_man;
-                       curp->outmdoc = html_mdoc;
-                       curp->outfree = html_free;
-                       break;
-               case (OUTT_TREE):
-                       curp->outman = tree_man;
-                       curp->outmdoc = tree_mdoc;
-                       break;
-               case (OUTT_ASCII):
-                       /* FALLTHROUGH */
-               case (OUTT_PS):
-                       curp->outman = terminal_man;
-                       curp->outmdoc = terminal_mdoc;
-                       break;
-               default:
-                       break;
+                       if ( ! rc) {
+                               assert(MANDOCLEVEL_FATAL <= exit_status);
+                               break;
+                       }
                }
                }
-       }
 
 
-       /* Execute the out device, if it exists. */
+               /* Temporary buffers typically are not full. */
 
 
-       if (man && curp->outman)
-               (*curp->outman)(curp->outdata, man);
-       if (mdoc && curp->outmdoc)
-               (*curp->outmdoc)(curp->outdata, mdoc);
+               if (0 == start && '\0' == blk.buf[i])
+                       break;
 
 
- cleanup:
-       memset(&curp->regs, 0, sizeof(struct regset));
-       if (mdoc)
-               mdoc_reset(mdoc);
-       if (man)
-               man_reset(man);
-       if (roff)
-               roff_reset(roff);
-       if (ln.buf)
-               free(ln.buf);
-       if (with_mmap)
-               munmap(blk.buf, blk.sz);
-       else
-               free(blk.buf);
+               /* Start the next input line. */
 
 
-       return;
+               pos = 0;
+       }
 
 
- bailout:
-       with_fatal = 1;
-       goto cleanup;
+       free(ln.buf);
 }
 
 }
 
-
-static int
-pset(const char *buf, int pos, struct curparse *curp,
-               struct man **man, struct mdoc **mdoc)
+static void
+pset(const char *buf, int pos, struct curparse *curp)
 {
        int              i;
 
 {
        int              i;
 
@@ -682,49 +826,53 @@ pset(const char *buf, int pos, struct curparse *curp,
         * explicitly.  If passed as -mandoc, then try to guess from the
         * line: either skip dot-lines, use -mdoc when finding `.Dt', or
         * default to -man, which is more lenient.
         * explicitly.  If passed as -mandoc, then try to guess from the
         * line: either skip dot-lines, use -mdoc when finding `.Dt', or
         * default to -man, which is more lenient.
+        *
+        * Separate out pmdoc/pman from mdoc/man: the first persists
+        * through all parsers, while the latter is used per-parse.
         */
 
        if ('.' == buf[0] || '\'' == buf[0]) {
                for (i = 1; buf[i]; i++)
                        if (' ' != buf[i] && '\t' != buf[i])
                                break;
         */
 
        if ('.' == buf[0] || '\'' == buf[0]) {
                for (i = 1; buf[i]; i++)
                        if (' ' != buf[i] && '\t' != buf[i])
                                break;
-               if (0 == buf[i])
-                       return(1);
+               if ('\0' == buf[i])
+                       return;
        }
 
        switch (curp->inttype) {
        case (INTT_MDOC):
        }
 
        switch (curp->inttype) {
        case (INTT_MDOC):
-               if (NULL == curp->mdoc) 
-                       curp->mdoc = mdoc_init(curp);
-               if (NULL == (*mdoc = curp->mdoc))
-                       return(0);
-               return(1);
+               if (NULL == curp->pmdoc) 
+                       curp->pmdoc = mdoc_alloc
+                               (&curp->regs, curp, mmsg);
+               assert(curp->pmdoc);
+               curp->mdoc = curp->pmdoc;
+               return;
        case (INTT_MAN):
        case (INTT_MAN):
-               if (NULL == curp->man) 
-                       curp->man = man_init(curp);
-               if (NULL == (*man = curp->man))
-                       return(0);
-               return(1);
+               if (NULL == curp->pman) 
+                       curp->pman = man_alloc
+                               (&curp->regs, curp, mmsg);
+               assert(curp->pman);
+               curp->man = curp->pman;
+               return;
        default:
                break;
        }
 
        if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3))  {
        default:
                break;
        }
 
        if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3))  {
-               if (NULL == curp->mdoc) 
-                       curp->mdoc = mdoc_init(curp);
-               if (NULL == (*mdoc = curp->mdoc))
-                       return(0);
-               return(1);
+               if (NULL == curp->pmdoc) 
+                       curp->pmdoc = mdoc_alloc
+                               (&curp->regs, curp, mmsg);
+               assert(curp->pmdoc);
+               curp->mdoc = curp->pmdoc;
+               return;
        } 
 
        } 
 
-       if (NULL == curp->man) 
-               curp->man = man_init(curp);
-       if (NULL == (*man = curp->man))
-               return(0);
-       return(1);
+       if (NULL == curp->pman) 
+               curp->pman = man_alloc(&curp->regs, curp, mmsg);
+       assert(curp->pman);
+       curp->man = curp->pman;
 }
 
 }
 
-
 static int
 moptions(enum intt *tflags, char *arg)
 {
 static int
 moptions(enum intt *tflags, char *arg)
 {
@@ -743,7 +891,6 @@ moptions(enum intt *tflags, char *arg)
        return(1);
 }
 
        return(1);
 }
 
-
 static int
 toptions(struct curparse *curp, char *arg)
 {
 static int
 toptions(struct curparse *curp, char *arg)
 {
@@ -752,8 +899,7 @@ toptions(struct curparse *curp, char *arg)
                curp->outtype = OUTT_ASCII;
        else if (0 == strcmp(arg, "lint")) {
                curp->outtype = OUTT_LINT;
                curp->outtype = OUTT_ASCII;
        else if (0 == strcmp(arg, "lint")) {
                curp->outtype = OUTT_LINT;
-               curp->wflags |= WARN_WALL;
-               curp->fflags |= FL_STRICT;
+               curp->wlevel  = MANDOCLEVEL_WARNING;
        }
        else if (0 == strcmp(arg, "tree"))
                curp->outtype = OUTT_TREE;
        }
        else if (0 == strcmp(arg, "tree"))
                curp->outtype = OUTT_TREE;
@@ -763,6 +909,8 @@ toptions(struct curparse *curp, char *arg)
                curp->outtype = OUTT_XHTML;
        else if (0 == strcmp(arg, "ps"))
                curp->outtype = OUTT_PS;
                curp->outtype = OUTT_XHTML;
        else if (0 == strcmp(arg, "ps"))
                curp->outtype = OUTT_PS;
+       else if (0 == strcmp(arg, "pdf"))
+               curp->outtype = OUTT_PDF;
        else {
                fprintf(stderr, "%s: Bad argument\n", arg);
                return(0);
        else {
                fprintf(stderr, "%s: Bad argument\n", arg);
                return(0);
@@ -771,44 +919,38 @@ toptions(struct curparse *curp, char *arg)
        return(1);
 }
 
        return(1);
 }
 
-
 static int
 static int
-foptions(int *fflags, char *arg)
+woptions(struct curparse *curp, char *arg)
 {
        char            *v, *o;
 {
        char            *v, *o;
-       const char      *toks[8];
+       const char      *toks[6]; 
 
 
-       toks[0] = "ign-scope";
-       toks[1] = "no-ign-escape";
-       toks[2] = "no-ign-macro";
-       toks[3] = "ign-errors";
-       toks[4] = "strict";
-       toks[5] = "ign-escape";
-       toks[6] = NULL;
+       toks[0] = "stop";
+       toks[1] = "all";
+       toks[2] = "warning";
+       toks[3] = "error";
+       toks[4] = "fatal";
+       toks[5] = NULL;
 
        while (*arg) {
                o = arg;
                switch (getsubopt(&arg, UNCONST(toks), &v)) {
                case (0):
 
        while (*arg) {
                o = arg;
                switch (getsubopt(&arg, UNCONST(toks), &v)) {
                case (0):
-                       *fflags |= FL_IGN_SCOPE;
+                       curp->wstop = 1;
                        break;
                case (1):
                        break;
                case (1):
-                       *fflags |= FL_NIGN_ESCAPE;
-                       break;
+                       /* FALLTHROUGH */
                case (2):
                case (2):
-                       *fflags |= FL_NIGN_MACRO;
+                       curp->wlevel = MANDOCLEVEL_WARNING;
                        break;
                case (3):
                        break;
                case (3):
-                       *fflags |= FL_IGN_ERRORS;
+                       curp->wlevel = MANDOCLEVEL_ERROR;
                        break;
                case (4):
                        break;
                case (4):
-                       *fflags |= FL_STRICT;
-                       break;
-               case (5):
-                       *fflags &= ~FL_NIGN_ESCAPE;
+                       curp->wlevel = MANDOCLEVEL_FATAL;
                        break;
                default:
                        break;
                default:
-                       fprintf(stderr, "%s: Bad argument\n", o);
+                       fprintf(stderr, "-W%s: Bad argument\n", o);
                        return(0);
                }
        }
                        return(0);
                }
        }
@@ -816,71 +958,29 @@ foptions(int *fflags, char *arg)
        return(1);
 }
 
        return(1);
 }
 
-
-static int
-woptions(int *wflags, char *arg)
-{
-       char            *v, *o;
-       const char      *toks[3]; 
-
-       toks[0] = "all";
-       toks[1] = "error";
-       toks[2] = NULL;
-
-       while (*arg) {
-               o = arg;
-               switch (getsubopt(&arg, UNCONST(toks), &v)) {
-               case (0):
-                       *wflags |= WARN_WALL;
-                       break;
-               case (1):
-                       *wflags |= WARN_WERR;
-                       break;
-               default:
-                       fprintf(stderr, "%s: Bad argument\n", o);
-                       return(0);
-               }
-       }
-
-       return(1);
-}
-
-
 static int
 mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg)
 {
        struct curparse *cp;
 static int
 mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg)
 {
        struct curparse *cp;
-       const char *level;
-       int rc;
+       enum mandoclevel level;
+
+       level = MANDOCLEVEL_FATAL;
+       while (t < mandoclimits[level])
+               /* LINTED */
+               level--;
 
        cp = (struct curparse *)arg;
 
        cp = (struct curparse *)arg;
-       level = NULL;
-       rc = 1;
-
-       if (t >= MANDOCERR_FATAL) {
-               with_fatal = 1;
-               level = "FATAL";
-               rc = 0;
-       } else {
-               if ( ! (WARN_WALL & cp->wflags))
-                       return(1);
-               if (t >= MANDOCERR_ERROR) {
-                       with_error = 1;
-                       level = "ERROR";
-               }
-               if (WARN_WERR & cp->wflags) {
-                       with_fatal = 1;
-                       rc = 0;
-               }
-       }
+       if (level < cp->wlevel)
+               return(1);
 
 
-       fprintf(stderr, "%s:%d:%d:", cp->file, ln, col + 1);
-       if (level)
-               fprintf(stderr, " %s:", level);
-       fprintf(stderr, " %s", mandocerrs[t]);
+       fprintf(stderr, "%s:%d:%d: %s: %s",
+           cp->file, ln, col + 1, mandoclevels[level], mandocerrs[t]);
        if (msg)
                fprintf(stderr, ": %s", msg);
        fputc('\n', stderr);
 
        if (msg)
                fprintf(stderr, ": %s", msg);
        fputc('\n', stderr);
 
-       return(rc);
+       if (exit_status < level)
+               exit_status = level;
+       
+       return(level < MANDOCLEVEL_FATAL);
 }
 }