=================================================================== RCS file: /cvs/mandoc/main.c,v retrieving revision 1.65 retrieving revision 1.84 diff -u -p -r1.65 -r1.84 --- mandoc/main.c 2010/05/15 04:46:10 1.65 +++ mandoc/main.c 2010/06/07 10:52:44 1.84 @@ -1,4 +1,4 @@ -/* $Id: main.c,v 1.65 2010/05/15 04:46:10 kristaps Exp $ */ +/* $Id: main.c,v 1.84 2010/06/07 10:52:44 kristaps Exp $ */ /* * Copyright (c) 2008, 2009 Kristaps Dzonsons * @@ -18,6 +18,7 @@ #include "config.h" #endif +#include #include #include @@ -28,8 +29,10 @@ #include #include +#include "mandoc.h" #include "mdoc.h" #include "man.h" +#include "roff.h" #include "main.h" #define UNCONST(a) ((void *)(uintptr_t)(const void *)(a)) @@ -69,6 +72,7 @@ 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; @@ -76,45 +80,110 @@ struct curparse { #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. */ - enum intt inttype; /* Input parsers... */ - struct man *man; - struct mdoc *mdoc; - enum outt outtype; /* Output devices... */ - out_mdoc outmdoc; - out_man outman; - out_free outfree; - void *outdata; - char outopts[BUFSIZ]; +#define FL_STRICT FL_NIGN_ESCAPE | \ + FL_NIGN_MACRO /* ignore nothing */ + enum intt inttype; /* which parser to use */ + struct man *man; /* man parser */ + struct mdoc *mdoc; /* mdoc parser */ + struct roff *roff; /* roff parser (!NULL) */ + 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 */ }; -#define FL_STRICT FL_NIGN_ESCAPE | \ - FL_NIGN_MACRO +static const char * const mandocerrs[MANDOCERR_MAX] = { + "ok", + "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", + "bad escape sequence", + "unterminated quoted string", + "argument requires the width argument", + "superfluous width argument", + "bad date argument", + "bad width argument", + "unknown manual section", + "section not in conventional manual section", + "end of line whitespace", + "scope open on exit", + "NAME section must come first", + "bad Boolean value", + "child violates parent syntax", + "bad AT&T symbol", + "list type repeated", + "display type repeated", + "argument repeated", + "manual name not yet set", + "obsolete macro ignored", + "empty macro ignored", + "macro not allowed in body", + "macro not allowed in prologue", + "bad character", + "bad NAME section contents", + "no blank lines", + "no text in this context", + "bad comment style", + "unknown macro will be lost", + "line scope broken", + "scope broken", + "argument count wrong", + "request scope close w/none open", + "scope already open", + "macro requires line argument(s)", + "macro requires body argument(s)", + "macro requires argument(s)", + "no title in document", + "missing list type", + "line argument(s) will be lost", + "body argument(s) will be lost", + "column syntax is inconsistent", + "missing font type", + "missing display type", + "displays may not be nested", + "no scope to rewind: syntax violated", + "scope broken, syntax violated", + "line scope broken, syntax violated", + "argument count wrong, violates syntax", + "child violates parent syntax", + "argument count wrong, violates syntax", + "no document body", + "no document prologue", + "utsname system call failed", + "memory exhausted", +}; +static void fdesc(struct curparse *); +static void ffile(const char *, struct curparse *); static int foptions(int *, char *); -static int toptions(struct curparse *, char *); +static struct man *man_init(struct curparse *); +static struct mdoc *mdoc_init(struct curparse *); +static struct roff *roff_init(struct curparse *); 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 void ffile(struct buf *, struct buf *, - const char *, struct curparse *); -static void fdesc(struct buf *, struct buf *, - struct curparse *); +static int mmsg(enum mandocerr, void *, + int, int, const char *); 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 *); -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 const char *progname; -static int with_error, with_warning; +static int with_error; +static int with_warning; int main(int argc, char *argv[]) { int c; - struct buf ln, blk; struct curparse curp; progname = strrchr(argv[0], '/'); @@ -162,32 +231,32 @@ main(int argc, char *argv[]) argc -= optind; argv += optind; - memset(&ln, 0, sizeof(struct buf)); - memset(&blk, 0, sizeof(struct buf)); - if (NULL == *argv) { curp.file = ""; curp.fd = STDIN_FILENO; - fdesc(&blk, &ln, &curp); + fdesc(&curp); } while (*argv) { - ffile(&blk, &ln, *argv, &curp); + ffile(*argv, &curp); if (with_error && !(curp.fflags & FL_IGN_ERRORS)) 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.roff) + roff_free(curp.roff); - return((with_warning || with_error) ? EXIT_FAILURE : EXIT_SUCCESS ); + return((with_warning || with_error) ? + EXIT_FAILURE : EXIT_SUCCESS); } @@ -215,11 +284,7 @@ static struct man * man_init(struct curparse *curp) { int pflags; - struct man_cb mancb; - mancb.man_err = merr; - mancb.man_warn = mwarn; - /* Defaults from mandoc.1. */ pflags = MAN_IGN_MACRO | MAN_IGN_ESCAPE; @@ -229,19 +294,23 @@ man_init(struct curparse *curp) if (curp->fflags & FL_NIGN_ESCAPE) pflags &= ~MAN_IGN_ESCAPE; - return(man_alloc(curp, pflags, &mancb)); + return(man_alloc(curp, pflags, mmsg)); } +static struct roff * +roff_init(struct curparse *curp) +{ + + return(roff_alloc(mmsg, curp)); +} + + static struct mdoc * mdoc_init(struct curparse *curp) { int pflags; - struct mdoc_cb mdoccb; - mdoccb.mdoc_err = merr; - mdoccb.mdoc_warn = mwarn; - /* Defaults from mandoc.1. */ pflags = MDOC_IGN_MACRO | MDOC_IGN_ESCAPE; @@ -253,13 +322,12 @@ mdoc_init(struct curparse *curp) if (curp->fflags & FL_NIGN_MACRO) pflags &= ~MDOC_IGN_MACRO; - return(mdoc_alloc(curp, pflags, &mdoccb)); + return(mdoc_alloc(curp, pflags, mmsg)); } static void -ffile(struct buf *blk, struct buf *ln, - const char *file, struct curparse *curp) +ffile(const char *file, struct curparse *curp) { curp->file = file; @@ -269,134 +337,229 @@ ffile(struct buf *blk, struct buf *ln, return; } - fdesc(blk, ln, curp); + fdesc(curp); if (-1 == close(curp->fd)) perror(curp->file); } -static void -fdesc(struct buf *blk, struct buf *ln, struct curparse *curp) +static int +resize_buf(struct buf *buf, size_t initial) { - size_t sz; - ssize_t ssz; - struct stat st; - int j, i, pos, lnn, comment; - struct man *man; - struct mdoc *mdoc; + void *tmp; + size_t sz; - sz = BUFSIZ; - man = NULL; - mdoc = NULL; + if (buf->sz == 0) + sz = initial; + else + sz = 2 * buf->sz; + tmp = realloc(buf->buf, sz); + if (NULL == tmp) { + perror(NULL); + return(0); + } + buf->buf = tmp; + buf->sz = sz; + return(1); +} - /* - * 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. - */ +static int +read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap) +{ + struct stat st; + size_t off; + ssize_t ssz; + if (-1 == fstat(curp->fd, &st)) { perror(curp->file); with_error = 1; - return; + return(0); } - if ((size_t)st.st_blksize > sz) - sz = st.st_blksize; - if (sz > blk->sz) { - void *buf = realloc(blk->buf, sz); + /* + * 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 (NULL == buf) { - perror(NULL); + if (S_ISREG(st.st_mode)) { + if (st.st_size >= (1U << 31)) { + fprintf(stderr, "%s: input too large\n", + curp->file); with_error = 1; - return; + return(0); } - blk->buf = buf; - 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))) { + *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; + } + if (! resize_buf(fb, 65536)) + break; + } + 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); - goto bailout; - } else if (0 == ssz) break; + } + off += (size_t)ssz; + } - /* Parse the read block into partial or full lines. */ + free(fb->buf); + fb->buf = NULL; + with_error = 1; + 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) { - perror(NULL); - goto bailout; - } - } - if ('\n' != blk->buf[i]) { - if (comment) - continue; - ln->buf[pos++] = blk->buf[i]; +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; - /* Handle in-line `\"' comments. */ + man = NULL; + mdoc = NULL; + roff = NULL; + memset(&ln, 0, sizeof(struct buf)); - if (1 == pos || '\"' != ln->buf[pos - 1]) - continue; + /* + * Two buffers: ln and buf. buf is the input file and may be + * memory mapped. ln is a line buffer and grows on-demand. + */ - for (j = pos - 2; j >= 0; j--) - if ('\\' != ln->buf[j]) - break; + if ( ! read_whole_file(curp, &blk, &with_mmap)) + return; - if ( ! ((pos - 2 - j) % 2)) - continue; + if (NULL == curp->roff) + curp->roff = roff_init(curp); + if (NULL == (roff = curp->roff)) + goto bailout; - comment = 1; - pos -= 2; + for (i = 0, lnn = 1; i < (int)blk.sz;) { + pos = 0; + lnn_start = lnn; + while (i < (int)blk.sz) { + if ('\n' == blk.buf[i]) { + ++i; + ++lnn; + break; + } + /* Trailing backslash is like a plain character. */ + if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) { + if (pos >= (int)ln.sz) + if (! resize_buf(&ln, 256)) + goto bailout; + ln.buf[pos++] = blk.buf[i++]; + continue; + } + /* Found an escape and at least one other character. */ + if ('\n' == blk.buf[i + 1]) { + /* Escaped newlines are skipped over */ + i += 2; + ++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] != ' ') + if (ln.buf[pos - 1] != ' ') break; - if (pos > 2 && ln->buf[pos - 2] == '\\') + if (pos > 2 && ln.buf[pos - 2] == '\\') break; } - continue; - } - - /* Handle escaped `\\n' newlines. */ - - 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; - } + break; } + /* Some other escape sequence, copy and continue. */ + if (pos + 1 >= (int)ln.sz) + if (! resize_buf(&ln, 256)) + goto bailout; - ln->buf[pos] = 0; - lnn++; + ln.buf[pos++] = blk.buf[i++]; + ln.buf[pos++] = blk.buf[i++]; + } - /* If unset, assign parser in pset(). */ - - if ( ! (man || mdoc) && ! pset(ln->buf, - pos, curp, &man, &mdoc)) + if (pos >= (int)ln.sz) + if (! resize_buf(&ln, 256)) goto bailout; + ln.buf[pos] = '\0'; - pos = comment = 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. + */ - /* Pass down into parsers. */ + of = 0; + do { + re = roff_parseln(roff, lnn_start, + &ln.buf, &ln.sz, of, &of); + } while (ROFF_RERUN == re); - if (man && ! man_parseln(man, lnn, ln->buf)) + if (ROFF_IGN == re) + continue; + else if (ROFF_ERR == re) + goto bailout; + + /* + * 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 ( ! (man || mdoc)) + if ( ! pset(ln.buf + of, pos - of, curp, &man, &mdoc)) goto bailout; - if (mdoc && ! mdoc_parseln(mdoc, lnn, ln->buf)) - goto bailout; - } + + /* Lastly, push down into the parsers themselves. */ + + 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. */ @@ -406,10 +569,14 @@ fdesc(struct buf *blk, struct buf *ln, struct curparse 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). */ @@ -434,7 +601,7 @@ fdesc(struct buf *blk, struct buf *ln, struct curparse case (OUTT_LINT): break; default: - curp->outdata = ascii_alloc(); + curp->outdata = ascii_alloc(curp->outopts); curp->outman = terminal_man; curp->outmdoc = terminal_mdoc; curp->outfree = terminal_free; @@ -450,14 +617,19 @@ fdesc(struct buf *blk, struct buf *ln, struct curparse (*curp->outmdoc)(curp->outdata, mdoc); cleanup: - if (curp->mdoc) { - mdoc_free(curp->mdoc); - curp->mdoc = NULL; - } - if (curp->man) { - man_free(curp->man); - curp->man = NULL; - } + 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); + return; bailout: @@ -480,7 +652,7 @@ pset(const char *buf, int pos, struct curparse *curp, * default to -man, which is more lenient. */ - if (buf[0] == '.') { + if ('.' == buf[0] || '\'' == buf[0]) { for (i = 1; buf[i]; i++) if (' ' != buf[i] && '\t' != buf[i]) break; @@ -640,42 +812,34 @@ woptions(int *wflags, char *arg) } -/* ARGSUSED */ static int -merr(void *arg, int line, int col, const char *msg) +mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg) { - struct curparse *curp; + struct curparse *cp; - curp = (struct curparse *)arg; + cp = (struct curparse *)arg; - (void)fprintf(stderr, "%s:%d:%d: error: %s\n", - curp->file, line, col + 1, msg); + if (t <= MANDOCERR_ERROR) { + if ( ! (cp->wflags & WARN_WALL)) + return(1); + with_warning = 1; + } else + with_error = 1; - with_error = 1; + fprintf(stderr, "%s:%d:%d: %s", cp->file, + ln, col + 1, mandocerrs[t]); - return(0); -} + if (msg) + fprintf(stderr, ": %s", msg); + fputc('\n', stderr); -static int -mwarn(void *arg, int line, int col, const char *msg) -{ - struct curparse *curp; - - curp = (struct curparse *)arg; - - if ( ! (curp->wflags & WARN_WALL)) - return(1); - - (void)fprintf(stderr, "%s:%d:%d: warning: %s\n", - curp->file, line, col + 1, msg); - - with_warning = 1; - if (curp->wflags & WARN_WERR) { + /* This is superfluous, but whatever. */ + if (t > MANDOCERR_ERROR) + return(0); + if (cp->wflags & WARN_WERR) { with_error = 1; return(0); } - return(1); } -