=================================================================== RCS file: /cvs/mandoc/main.c,v retrieving revision 1.41 retrieving revision 1.135 diff -u -p -r1.41 -r1.135 --- mandoc/main.c 2009/07/28 10:15:12 1.41 +++ mandoc/main.c 2011/01/04 15:02:00 1.135 @@ -1,6 +1,7 @@ -/* $Id: main.c,v 1.41 2009/07/28 10:15:12 kristaps Exp $ */ +/* $Id: main.c,v 1.135 2011/01/04 15:02:00 kristaps Exp $ */ /* - * Copyright (c) 2008, 2009 Kristaps Dzonsons + * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons + * Copyright (c) 2010 Ingo Schwarze * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above @@ -14,34 +15,45 @@ * 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 #include #include -#include +#include #include #include +#include #include #include #include +#include "mandoc.h" +#include "main.h" #include "mdoc.h" #include "man.h" +#include "roff.h" -/* Account for FreeBSD and Linux in our declarations. */ +#ifndef MAP_FILE +#define MAP_FILE 0 +#endif -#ifdef __linux__ -extern int getsubopt(char **, char * const *, char **); -# ifndef __dead -# define __dead __attribute__((__noreturn__)) +#define REPARSE_LIMIT 1000 +#define UNCONST(a) ((void *)(uintptr_t)(const void *)(a)) + +/* FIXME: Intel's compiler? LLVM? pcc? */ + +#if !defined(__GNUC__) || (__GNUC__ < 2) +# if !defined(lint) +# define __attribute__(x) # endif -#elif defined(__dead2) -# ifndef __dead -# define __dead __dead2 -# endif -#endif +#endif /* !defined(__GNUC__) || (__GNUC__ < 2) */ -typedef int (*out_mdoc)(void *, const struct mdoc *); -typedef int (*out_man)(void *, const struct man *); +typedef void (*out_mdoc)(void *, const struct mdoc *); +typedef void (*out_man)(void *, const struct man *); typedef void (*out_free)(void *); struct buf { @@ -58,90 +70,214 @@ enum intt { enum outt { OUTT_ASCII = 0, OUTT_TREE, - OUTT_LINT + OUTT_HTML, + OUTT_XHTML, + OUTT_LINT, + OUTT_PS, + OUTT_PDF }; struct curparse { const char *file; /* Current parse. */ int fd; /* Current parse. */ - int wflags; -#define WARN_WALL (1 << 0) /* All-warnings mask. */ -#define WARN_WERR (1 << 2) /* Warnings->errors. */ - int fflags; -#define IGN_SCOPE (1 << 0) /* Ignore scope errors. */ -#define NO_IGN_ESCAPE (1 << 1) /* Don't ignore bad escapes. */ -#define NO_IGN_MACRO (1 << 2) /* Don't ignore bad macros. */ -#define NO_IGN_CHARS (1 << 3) /* Don't ignore bad chars. */ -#define IGN_ERRORS (1 << 4) /* Ignore failed parse. */ - enum intt inttype; /* Input parsers... */ - struct man *man; - struct man *lastman; - struct mdoc *mdoc; - struct mdoc *lastmdoc; - enum outt outtype; /* Output devices... */ - out_mdoc outmdoc; - out_man outman; - out_free outfree; - void *outdata; + 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 */ + 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 regset regs; /* roff registers */ + int reparse_count; /* finite interpolation stack */ + enum outt outtype; /* which output to use */ + out_mdoc outmdoc; /* mdoc output ptr */ + out_man outman; /* man output ptr */ + out_free outfree; /* free output ptr */ + void *outdata; /* data for output */ + char outopts[BUFSIZ]; /* buf of output opts */ }; -extern void *ascii_alloc(void); -extern int tree_mdoc(void *, const struct mdoc *); -extern int tree_man(void *, const struct man *); -extern int terminal_mdoc(void *, const struct mdoc *); -extern int terminal_man(void *, const struct man *); -extern void terminal_free(void *); +static const char * const mandoclevels[MANDOCLEVEL_MAX] = { + "SUCCESS", + "RESERVED", + "WARNING", + "ERROR", + "FATAL", + "BADARG", + "SYSERR" +}; -static int foptions(int *, char *); -static int toptions(enum outt *, char *); -static int moptions(enum intt *, char *); -static int woptions(int *, char *); -static int merr(void *, int, int, const char *); -static int mwarn(void *, int, int, const char *); -static int ffile(struct buf *, struct buf *, - const char *, struct curparse *); -static int fdesc(struct buf *, struct buf *, - struct curparse *); -static int pset(const char *, int, struct curparse *, - struct man **, struct mdoc **); -static struct man *man_init(struct curparse *); -static struct mdoc *mdoc_init(struct curparse *); -__dead static void version(void); -__dead static void usage(void); +static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = { + MANDOCERR_OK, + MANDOCERR_WARNING, + MANDOCERR_WARNING, + MANDOCERR_ERROR, + MANDOCERR_FATAL, + MANDOCERR_MAX, + MANDOCERR_MAX +}; -extern char *__progname; +static const char * const mandocerrs[MANDOCERR_MAX] = { + "ok", + "generic warning", + /* related to the prologue */ + "no title in document", + "document title should be all caps", + "unknown manual section", + "cannot parse date argument", + "prologue macros out of order", + "duplicate prologue macro", + "macro not allowed in prologue", + "macro not allowed in body", + + /* related to document structure */ + ".so is fragile, better use ln(1)", + "NAME section must come first", + "bad NAME section contents", + "manual name not yet set", + "sections out of conventional order", + "duplicate section name", + "section not in conventional manual section", + + /* related to macros and nesting */ + "skipping obsolete macro", + "skipping paragraph macro", + "blocks badly nested", + "child violates parent syntax", + "nested displays are not portable", + "already in literal mode", + + /* related to missing macro arguments */ + "skipping empty macro", + "argument count wrong", + "missing display type", + "list type must come first", + "tag lists require a width argument", + "missing font type", + + /* related to bad macro arguments */ + "skipping argument", + "duplicate argument", + "duplicate display type", + "duplicate list type", + "unknown AT&T UNIX version", + "bad Boolean value", + "unknown font", + "unknown standard specifier", + "bad width argument", + + /* related to plain text */ + "blank line in non-literal context", + "tab in non-literal context", + "end of line whitespace", + "bad comment style", + "unknown escape sequence", + "unterminated quoted string", + + /* related to tables */ + "extra data cells", + + "generic error", + + /* related to tables */ + "bad table syntax", + "bad table option", + "bad table layout", + "no table layout cells specified", + "no table data cells specified", + "ignore data in cell", + "data block still open", + + "input stack limit exceeded, infinite loop?", + "skipping bad character", + "skipping text before the first section header", + "skipping unknown macro", + "NOT IMPLEMENTED: skipping request", + "line scope broken", + "argument count wrong", + "skipping end of block that is not open", + "missing end of block", + "scope open on exit", + "uname(3) system call failed", + "macro requires line argument(s)", + "macro requires body argument(s)", + "macro requires argument(s)", + "missing list type", + "line argument(s) will be lost", + "body argument(s) will be lost", + + "generic fatal error", + + "column syntax is inconsistent", + "NOT IMPLEMENTED: .Bd -file", + "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", + "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 int pfile(const char *, struct curparse *); +static int moptions(enum intt *, char *); +static int mmsg(enum mandocerr, void *, + int, int, const char *); +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 woptions(struct curparse *, char *); + +static const char *progname; +static enum mandoclevel file_status = MANDOCLEVEL_OK; +static enum mandoclevel exit_status = MANDOCLEVEL_OK; + int main(int argc, char *argv[]) { - int c, rc; - struct buf ln, blk; + int c; struct curparse curp; - bzero(&curp, sizeof(struct curparse)); + progname = strrchr(argv[0], '/'); + if (progname == NULL) + progname = argv[0]; + else + ++progname; + memset(&curp, 0, sizeof(struct curparse)); + curp.inttype = INTT_AUTO; curp.outtype = OUTT_ASCII; + curp.wlevel = MANDOCLEVEL_FATAL; /* LINTED */ - while (-1 != (c = getopt(argc, argv, "f:m:VW:T:"))) + while (-1 != (c = getopt(argc, argv, "m:O:T:VW:"))) switch (c) { - case ('f'): - if ( ! foptions(&curp.fflags, optarg)) - return(EXIT_FAILURE); - break; case ('m'): if ( ! moptions(&curp.inttype, optarg)) - return(EXIT_FAILURE); + return((int)MANDOCLEVEL_BADARG); break; + case ('O'): + (void)strlcat(curp.outopts, optarg, BUFSIZ); + (void)strlcat(curp.outopts, ",", BUFSIZ); + break; case ('T'): - if ( ! toptions(&curp.outtype, optarg)) - return(EXIT_FAILURE); + if ( ! toptions(&curp, optarg)) + return((int)MANDOCLEVEL_BADARG); break; case ('W'): - if ( ! woptions(&curp.wflags, optarg)) - return(EXIT_FAILURE); + if ( ! woptions(&curp, optarg)) + return((int)MANDOCLEVEL_BADARG); break; case ('V'): version(); @@ -154,316 +290,588 @@ main(int argc, char *argv[]) argc -= optind; argv += optind; - bzero(&ln, sizeof(struct buf)); - bzero(&blk, sizeof(struct buf)); - - rc = 1; - if (NULL == *argv) { curp.file = ""; curp.fd = STDIN_FILENO; - c = fdesc(&blk, &ln, &curp); - if ( ! (IGN_ERRORS & curp.fflags)) - rc = 1 == c ? 1 : 0; - else - rc = -1 == c ? 0 : 1; + fdesc(&curp); } - while (rc && *argv) { - c = ffile(&blk, &ln, *argv, &curp); - if ( ! (IGN_ERRORS & curp.fflags)) - rc = 1 == c ? 1 : 0; - else - rc = -1 == c ? 0 : 1; - - argv++; - if (*argv && rc) { - if (curp.lastman) - if ( ! man_reset(curp.lastman)) - rc = 0; - if (curp.lastmdoc) - if ( ! mdoc_reset(curp.lastmdoc)) - rc = 0; - curp.lastman = NULL; - curp.lastmdoc = NULL; - } + while (*argv) { + ffile(*argv, &curp); + if (MANDOCLEVEL_OK != exit_status && curp.wstop) + break; + ++argv; } - if (blk.buf) - free(blk.buf); - if (ln.buf) - free(ln.buf); 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); - return(rc ? EXIT_SUCCESS : EXIT_FAILURE); + return((int)exit_status); } -__dead static void +static void version(void) { - (void)printf("%s %s\n", __progname, VERSION); - exit(EXIT_SUCCESS); + (void)printf("%s %s\n", progname, VERSION); + exit((int)MANDOCLEVEL_OK); } -__dead static void +static void usage(void) { - (void)fprintf(stderr, "usage: %s [-V] [-foption...] " - "[-mformat] [-Toutput] [-Werr...]\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; - struct man *man; - struct man_cb mancb; - mancb.man_err = merr; - mancb.man_warn = mwarn; + /* + * 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(). + */ - /* Defaults from mandoc.1. */ + curp->file = file; - pflags = MAN_IGN_MACRO | MAN_IGN_ESCAPE | MAN_IGN_CHARS; + if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) { + perror(curp->file); + exit_status = MANDOCLEVEL_SYSERR; + return; + } - if (curp->fflags & NO_IGN_MACRO) - pflags &= ~MAN_IGN_MACRO; - if (curp->fflags & NO_IGN_CHARS) - pflags &= ~MAN_IGN_CHARS; - if (curp->fflags & NO_IGN_ESCAPE) - pflags &= ~MAN_IGN_ESCAPE; + fdesc(curp); - if (NULL == (man = man_alloc(curp, pflags, &mancb))) - warnx("memory exhausted"); - - return(man); + 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; - struct mdoc *mdoc; - struct mdoc_cb mdoccb; + const char *savefile; + int fd, savefd; - mdoccb.mdoc_err = merr; - mdoccb.mdoc_warn = mwarn; + if (-1 == (fd = open(file, O_RDONLY, 0))) { + perror(file); + file_status = MANDOCLEVEL_SYSERR; + return(0); + } - /* Defaults from mandoc.1. */ + savefile = curp->file; + savefd = curp->fd; - pflags = MDOC_IGN_MACRO | MDOC_IGN_ESCAPE | MDOC_IGN_CHARS; + curp->file = file; + curp->fd = fd; - if (curp->fflags & IGN_SCOPE) - pflags |= MDOC_IGN_SCOPE; - if (curp->fflags & NO_IGN_ESCAPE) - pflags &= ~MDOC_IGN_ESCAPE; - if (curp->fflags & NO_IGN_MACRO) - pflags &= ~MDOC_IGN_MACRO; - if (curp->fflags & NO_IGN_CHARS) - pflags &= ~MDOC_IGN_CHARS; + pdesc(curp); - if (NULL == (mdoc = mdoc_alloc(curp, pflags, &mdoccb))) - warnx("memory exhausted"); + curp->file = savefile; + curp->fd = savefd; - return(mdoc); + if (-1 == close(fd)) + perror(file); + + return(MANDOCLEVEL_FATAL > file_status ? 1 : 0); } -static int -ffile(struct buf *blk, struct buf *ln, - const char *file, struct curparse *curp) +static void +resize_buf(struct buf *buf, size_t initial) { - int c; - curp->file = file; - if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) { - warn("%s", curp->file); - return(-1); + buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial; + buf->buf = realloc(buf->buf, buf->sz); + if (NULL == buf->buf) { + perror(NULL); + exit((int)MANDOCLEVEL_SYSERR); } - - c = fdesc(blk, ln, curp); - - if (-1 == close(curp->fd)) - warn("%s", curp->file); - - return(c); } static int -fdesc(struct buf *blk, struct buf *ln, struct curparse *curp) +read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap) { - size_t sz; - ssize_t ssz; struct stat st; - int j, i, pos, lnn, comment; - struct man *man; - struct mdoc *mdoc; + size_t off; + ssize_t ssz; - sz = BUFSIZ; - man = NULL; - mdoc = NULL; + if (-1 == fstat(curp->fd, &st)) { + perror(curp->file); + return(0); + } /* - * Two buffers: ln and buf. buf is the input buffer optimised - * here for each file's block size. ln is a line buffer. Both - * growable, hence passed in by ptr-ptr. + * If we're a regular file, try just reading in the whole entry + * via mmap(). This is faster than reading it into blocks, and + * since each file is only a few bytes to begin with, I'm not + * concerned that this is going to tank any machines. */ - if (-1 == fstat(curp->fd, &st)) - warn("%s", curp->file); - else if ((size_t)st.st_blksize > sz) - sz = st.st_blksize; - - if (sz > blk->sz) { - blk->buf = realloc(blk->buf, sz); - if (NULL == blk->buf) { - warn("realloc"); - return(-1); + if (S_ISREG(st.st_mode)) { + if (st.st_size >= (1U << 31)) { + fprintf(stderr, "%s: input too large\n", + curp->file); + return(0); } - blk->sz = sz; + *with_mmap = 1; + fb->sz = (size_t)st.st_size; + fb->buf = mmap(NULL, fb->sz, PROT_READ, + MAP_FILE|MAP_SHARED, curp->fd, 0); + if (fb->buf != MAP_FAILED) + return(1); } - /* Fill buf with file blocksize. */ + /* + * If this isn't a regular file (like, say, stdin), then we must + * go the old way and just read things in bit by bit. + */ - for (lnn = pos = comment = 0; ; ) { - if (-1 == (ssz = read(curp->fd, blk->buf, sz))) { - warn("%s", curp->file); - return(-1); - } else if (0 == ssz) + *with_mmap = 0; + off = 0; + fb->sz = 0; + fb->buf = NULL; + for (;;) { + if (off == fb->sz) { + if (fb->sz == (1U << 31)) { + fprintf(stderr, "%s: input too large\n", + curp->file); + break; + } + resize_buf(fb, 65536); + } + ssz = read(curp->fd, fb->buf + (int)off, fb->sz - off); + if (ssz == 0) { + fb->sz = off; + return(1); + } + if (ssz == -1) { + perror(curp->file); break; + } + off += (size_t)ssz; + } - /* Parse the read block into partial or full lines. */ + free(fb->buf); + fb->buf = NULL; + return(0); +} - for (i = 0; i < (int)ssz; i++) { - if (pos >= (int)ln->sz) { - ln->sz += 256; /* Step-size. */ - ln->buf = realloc(ln->buf, ln->sz); - if (NULL == ln->buf) { - warn("realloc"); - return(-1); - } - } - if ('\n' != blk->buf[i]) { - if (comment) - continue; - ln->buf[pos++] = blk->buf[i]; +static void +fdesc(struct curparse *curp) +{ - /* Handle in-line `\"' comments. */ + /* + * 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. + */ - if (1 == pos || '\"' != ln->buf[pos - 1]) - continue; + /* Zero the parse type. */ - for (j = pos - 2; j >= 0; j--) - if ('\\' != ln->buf[j]) - break; + curp->mdoc = NULL; + curp->man = NULL; + file_status = MANDOCLEVEL_OK; - if ( ! ((pos - 2 - j) % 2)) - continue; + /* Make sure the mandotory roff parser is initialised. */ - comment = 1; - pos -= 2; - continue; - } + if (NULL == curp->roff) { + curp->roff = roff_alloc(&curp->regs, curp, mmsg); + assert(curp->roff); + } - /* Handle escaped `\\n' newlines. */ + /* Fully parse the file. */ - if (pos > 0 && 0 == comment && - '\\' == ln->buf[pos - 1]) { - for (j = pos - 1; j >= 0; j--) - if ('\\' != ln->buf[j]) - break; - if ( ! ((pos - j) % 2)) { - pos--; - lnn++; - continue; - } - } + pdesc(curp); - ln->buf[pos] = 0; - lnn++; + if (MANDOCLEVEL_FATAL <= file_status) + goto cleanup; - /* If unset, assign parser in pset(). */ + /* NOTE a parser may not have been assigned, yet. */ - if ( ! (man || mdoc) && ! pset(ln->buf, - pos, curp, &man, &mdoc)) - return(-1); + if ( ! (curp->man || curp->mdoc)) { + fprintf(stderr, "%s: Not a manual\n", curp->file); + file_status = MANDOCLEVEL_FATAL; + goto cleanup; + } - pos = comment = 0; + /* Clean up the parse routine ASTs. */ - /* Pass down into parsers. */ + if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) { + assert(MANDOCLEVEL_FATAL <= file_status); + goto cleanup; + } - if (man && ! man_parseln(man, lnn, ln->buf)) - return(0); - if (mdoc && ! mdoc_parseln(mdoc, lnn, ln->buf)) - return(0); - } + if (curp->man && ! man_endparse(curp->man)) { + assert(MANDOCLEVEL_FATAL <= file_status); + goto cleanup; } - /* NOTE a parser may not have been assigned, yet. */ + assert(curp->roff); + roff_endparse(curp->roff); - if ( ! (man || mdoc)) { - (void)fprintf(stderr, "%s: not a manual\n", - curp->file); - return(0); - } + /* + * With -Wstop and warnings or errors of at least + * the requested level, do not produce output. + */ - if (mdoc && ! mdoc_endparse(mdoc)) - return(0); - if (man && ! man_endparse(man)) - return(0); + if (MANDOCLEVEL_OK != file_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_LINT): - break; - default: - curp->outdata = ascii_alloc(); + case (OUTT_PDF): + /* FALLTHROUGH */ + case (OUTT_ASCII): + /* FALLTHROUGH */ + case (OUTT_PS): curp->outman = terminal_man; curp->outmdoc = terminal_mdoc; - curp->outfree = terminal_free; break; + default: + break; } } /* Execute the out device, if it exists. */ - if (man && curp->outman) - if ( ! (*curp->outman)(curp->outdata, man)) - return(-1); - if (mdoc && curp->outmdoc) - if ( ! (*curp->outmdoc)(curp->outdata, mdoc)) - return(-1); + if (curp->man && curp->outman) + (*curp->outman)(curp->outdata, curp->man); + if (curp->mdoc && curp->outmdoc) + (*curp->outmdoc)(curp->outdata, curp->mdoc); - return(1); + 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); + + if (exit_status < file_status) + exit_status = file_status; + + return; } +static void +pdesc(struct curparse *curp) +{ + struct buf blk; + int with_mmap; -static int -pset(const char *buf, int pos, struct curparse *curp, - struct man **man, struct mdoc **mdoc) + /* + * 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)) { + file_status = MANDOCLEVEL_SYSERR; + return; + } + + /* Line number is per-file. */ + + 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; + enum rofferr rr; + 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; + curp->reparse_count = 0; + } + + while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) { + if ('\n' == blk.buf[i]) { + ++i; + ++lnn; + break; + } + + /* + * 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) + resize_buf(&ln, 256); + ln.buf[pos++] = blk.buf[i++]; + continue; + } + + /* Found escape & at least one other char. */ + + if ('\n' == blk.buf[i + 1]) { + i += 2; + /* Escaped newlines are skipped over */ + ++lnn; + continue; + } + + if ('"' == blk.buf[i + 1]) { + i += 2; + /* Comment, skip to end of line */ + for (; i < (int)blk.sz; ++i) { + if ('\n' == blk.buf[i]) { + ++i; + ++lnn; + break; + } + } + + /* Backout trailing whitespaces */ + for (; pos > 0; --pos) { + if (ln.buf[pos - 1] != ' ') + break; + if (pos > 2 && ln.buf[pos - 2] == '\\') + break; + } + break; + } + + /* Some other escape sequence, copy & cont. */ + + if (pos + 1 >= (int)ln.sz) + resize_buf(&ln, 256); + + ln.buf[pos++] = blk.buf[i++]; + ln.buf[pos++] = blk.buf[i++]; + } + + if (pos >= (int)ln.sz) + resize_buf(&ln, 256); + + ln.buf[pos] = '\0'; + + /* + * A significant amount of complexity is contained by + * the roff preprocessor. It's line-oriented but can be + * expressed on one line, so we need at times to + * readjust our starting point and re-run it. The roff + * preprocessor can also readjust the buffers with new + * data, so we pass them in wholesale. + */ + + of = 0; + +rerun: + rr = roff_parseln + (curp->roff, curp->line, + &ln.buf, &ln.sz, of, &of); + + switch (rr) { + case (ROFF_REPARSE): + if (REPARSE_LIMIT >= ++curp->reparse_count) + parsebuf(curp, ln, 0); + else + mmsg(MANDOCERR_ROFFLOOP, curp, + curp->line, pos, NULL); + pos = 0; + continue; + case (ROFF_APPEND): + pos = strlen(ln.buf); + continue; + case (ROFF_RERUN): + goto rerun; + case (ROFF_IGN): + pos = 0; + continue; + case (ROFF_ERR): + assert(MANDOCLEVEL_FATAL <= file_status); + break; + case (ROFF_SO): + if (pfile(ln.buf + of, curp)) { + pos = 0; + continue; + } else + break; + default: + break; + } + + /* + * If we encounter errors in the recursive parsebuf() + * call, make sure we don't continue parsing. + */ + + if (MANDOCLEVEL_FATAL <= file_status) + break; + + /* + * If input parsers have not been allocated, do so now. + * We keep these instanced betwen parsers, but set them + * locally per parse routine since we can use different + * parsers with each one. + */ + + if ( ! (curp->man || curp->mdoc)) + pset(ln.buf + of, pos - of, curp); + + /* + * Lastly, push down into the parsers themselves. One + * of these will have already been set in the pset() + * routine. + * If libroff returns ROFF_TBL, then add it to the + * currently open parse. Since we only get here if + * there does exist data (see tbl_data.c), we're + * guaranteed that something's been allocated. + */ + + if (ROFF_TBL == rr) { + assert(curp->man || curp->mdoc); + if (curp->man) + man_addspan(curp->man, roff_span(curp->roff)); + else + mdoc_addspan(curp->mdoc, roff_span(curp->roff)); + + } else 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); + + if ( ! rc) { + assert(MANDOCLEVEL_FATAL <= file_status); + break; + } + } + + /* Temporary buffers typically are not full. */ + + if (0 == start && '\0' == blk.buf[i]) + break; + + /* Start the next input line. */ + + pos = 0; + } + + free(ln.buf); +} + +static void +pset(const char *buf, int pos, struct curparse *curp) +{ int i; /* @@ -472,53 +880,53 @@ 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. + * + * Separate out pmdoc/pman from mdoc/man: the first persists + * through all parsers, while the latter is used per-parse. */ - if (buf[0] == '.') { + 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): - if (NULL == curp->mdoc) - curp->mdoc = mdoc_init(curp); - if (NULL == (*mdoc = curp->mdoc)) - return(0); - curp->lastmdoc = *mdoc; - 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): - if (NULL == curp->man) - curp->man = man_init(curp); - if (NULL == (*man = curp->man)) - return(0); - curp->lastman = *man; - 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)) { - if (NULL == curp->mdoc) - curp->mdoc = mdoc_init(curp); - if (NULL == (*mdoc = curp->mdoc)) - return(0); - curp->lastmdoc = *mdoc; - 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); - curp->lastman = *man; - 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) { @@ -530,71 +938,73 @@ moptions(enum intt *tflags, char *arg) else if (0 == strcmp(arg, "an")) *tflags = INTT_MAN; else { - warnx("bad argument: -m%s", arg); + fprintf(stderr, "%s: Bad argument\n", arg); return(0); } return(1); } - static int -toptions(enum outt *tflags, char *arg) +toptions(struct curparse *curp, char *arg) { if (0 == strcmp(arg, "ascii")) - *tflags = OUTT_ASCII; - else if (0 == strcmp(arg, "lint")) - *tflags = OUTT_LINT; + curp->outtype = OUTT_ASCII; + else if (0 == strcmp(arg, "lint")) { + curp->outtype = OUTT_LINT; + curp->wlevel = MANDOCLEVEL_WARNING; + } else if (0 == strcmp(arg, "tree")) - *tflags = OUTT_TREE; + curp->outtype = OUTT_TREE; + else if (0 == strcmp(arg, "html")) + curp->outtype = OUTT_HTML; + else if (0 == strcmp(arg, "xhtml")) + 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 { - warnx("bad argument: -T%s", arg); + fprintf(stderr, "%s: Bad argument\n", arg); return(0); } return(1); } - static int -foptions(int *fflags, char *arg) +woptions(struct curparse *curp, char *arg) { char *v, *o; - char *toks[7]; + const char *toks[6]; - toks[0] = "ign-scope"; - toks[1] = "no-ign-escape"; - toks[2] = "no-ign-macro"; - toks[3] = "no-ign-chars"; - toks[4] = "ign-errors"; - toks[5] = "strict"; - 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, toks, &v)) { + switch (getsubopt(&arg, UNCONST(toks), &v)) { case (0): - *fflags |= IGN_SCOPE; + curp->wstop = 1; break; case (1): - *fflags |= NO_IGN_ESCAPE; - break; + /* FALLTHROUGH */ case (2): - *fflags |= NO_IGN_MACRO; + curp->wlevel = MANDOCLEVEL_WARNING; break; case (3): - *fflags |= NO_IGN_CHARS; + curp->wlevel = MANDOCLEVEL_ERROR; break; case (4): - *fflags |= IGN_ERRORS; + curp->wlevel = MANDOCLEVEL_FATAL; break; - case (5): - *fflags |= NO_IGN_ESCAPE | - NO_IGN_MACRO | NO_IGN_CHARS; - break; default: - warnx("bad argument: -f%s", o); + fprintf(stderr, "-W%s: Bad argument\n", o); return(0); } } @@ -602,67 +1012,29 @@ foptions(int *fflags, char *arg) return(1); } - static int -woptions(int *wflags, char *arg) +mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg) { - char *v, *o; - char *toks[3]; + struct curparse *cp; + enum mandoclevel level; - toks[0] = "all"; - toks[1] = "error"; - toks[2] = NULL; + level = MANDOCLEVEL_FATAL; + while (t < mandoclimits[level]) + /* LINTED */ + level--; - while (*arg) { - o = arg; - switch (getsubopt(&arg, toks, &v)) { - case (0): - *wflags |= WARN_WALL; - break; - case (1): - *wflags |= WARN_WERR; - break; - default: - warnx("bad argument: -W%s", o); - return(0); - } - } - - return(1); -} - - -/* ARGSUSED */ -static int -merr(void *arg, int line, int col, const char *msg) -{ - struct curparse *curp; - - curp = (struct curparse *)arg; - - (void)fprintf(stderr, "%s:%d:%d: error: %s\n", - curp->file, line, col + 1, msg); - - return(0); -} - - -static int -mwarn(void *arg, int line, int col, const char *msg) -{ - struct curparse *curp; - - curp = (struct curparse *)arg; - - if ( ! (curp->wflags & WARN_WALL)) + cp = (struct curparse *)arg; + if (level < cp->wlevel) return(1); - (void)fprintf(stderr, "%s:%d:%d: warning: %s\n", - curp->file, line, col + 1, msg); + 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 ( ! (curp->wflags & WARN_WERR)) - return(1); + if (file_status < level) + file_status = level; - return(0); + return(level < MANDOCLEVEL_FATAL); } -