=================================================================== RCS file: /cvs/mandoc/main.c,v retrieving revision 1.111 retrieving revision 1.127 diff -u -p -r1.111 -r1.127 --- mandoc/main.c 2010/12/01 15:09:01 1.111 +++ mandoc/main.c 2010/12/29 14:38:14 1.127 @@ -1,4 +1,4 @@ -/* $Id: main.c,v 1.111 2010/12/01 15:09:01 kristaps Exp $ */ +/* $Id: main.c,v 1.127 2010/12/29 14:38:14 kristaps Exp $ */ /* * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons * Copyright (c) 2010 Ingo Schwarze @@ -41,6 +41,7 @@ #define MAP_FILE 0 #endif +#define REPARSE_LIMIT 1000 #define UNCONST(a) ((void *)(uintptr_t)(const void *)(a)) /* FIXME: Intel's compiler? LLVM? pcc? */ @@ -83,10 +84,13 @@ struct curparse { 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 */ @@ -120,77 +124,94 @@ static const char * const mandocerrs[MANDOCERR_MAX] = "generic warning", - "text should be uppercase", - "sections out of conventional order", - "section name repeats", - "out of order prologue", - "repeated prologue entry", - "list type must come first", - "tab in non-literal context", - "bad escape sequence", - "unterminated quoted string", - "argument requires the width argument", - "superfluous width argument", - "bad date argument", - "bad width argument", + /* 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", - "end of line whitespace", + + /* 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", - "generic error", + /* related to missing macro arguments */ + "skipping empty macro", + "missing display type", + "list type must come first", + "tag lists require a width argument", + "missing font type", - "NAME section must come first", + /* related to bad macro arguments */ + "skipping argument", + "duplicate argument", + "duplicate display type", + "duplicate list type", + "unknown AT&T UNIX version", "bad Boolean value", - "child violates parent syntax", - "bad AT&T symbol", - "bad standard", - "list type repeated", - "display type repeated", - "argument repeated", - "ignoring argument", - "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", + "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 macro will be lost", + "unknown escape sequence", + "unterminated quoted string", + + "generic error", + + "bad table syntax", + "bad table option", + "bad table layout", + "no table layout cells specified", + "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", - "request scope close w/none open", - "scope already open", + "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)", - "no title in document", "missing list type", - "missing display type", - "missing font type", "line argument(s) will be lost", "body argument(s) will be lost", - "paragraph macro ignored", "generic fatal error", "column syntax is inconsistent", - "displays may not be nested", - "unsupported display type", - "blocks badly nested", - "no such block is open", + "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", - "utsname system call failed", "static buffer exhausted", }; @@ -202,14 +223,14 @@ 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 *, - 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 woptions(struct curparse *, char *); static const char *progname; +static enum mandoclevel file_status = MANDOCLEVEL_OK; static enum mandoclevel exit_status = MANDOCLEVEL_OK; int @@ -276,10 +297,10 @@ main(int argc, char *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); @@ -300,18 +321,32 @@ static void usage(void) { - (void)fprintf(stderr, "usage: %s [-V] [-foption] " - "[-mformat] [-Ooption] [-Toutput] " - "[-Werr] [file...]\n", progname); + (void)fprintf(stderr, "usage: %s " + "[-V] " + "[-foption] " + "[-mformat] " + "[-Ooption] " + "[-Toutput] " + "[-Werr] " + "[file...]\n", + progname); + exit((int)MANDOCLEVEL_BADARG); } - static void ffile(const char *file, struct curparse *curp) { + /* + * 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(). + */ + curp->file = file; + if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) { perror(curp->file); exit_status = MANDOCLEVEL_SYSERR; @@ -332,7 +367,7 @@ pfile(const char *file, struct curparse *curp) if (-1 == (fd = open(file, O_RDONLY, 0))) { perror(file); - exit_status = MANDOCLEVEL_SYSERR; + file_status = MANDOCLEVEL_SYSERR; return(0); } @@ -350,7 +385,7 @@ pfile(const char *file, struct curparse *curp) if (-1 == close(fd)) perror(file); - return(MANDOCLEVEL_FATAL > exit_status ? 1 : 0); + return(MANDOCLEVEL_FATAL > file_status ? 1 : 0); } @@ -358,7 +393,7 @@ static void resize_buf(struct buf *buf, size_t initial) { - buf->sz = buf->sz ? 2 * buf->sz : initial; + buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial; buf->buf = realloc(buf->buf, buf->sz); if (NULL == buf->buf) { perror(NULL); @@ -439,39 +474,60 @@ read_whole_file(struct curparse *curp, struct buf *fb, static void fdesc(struct curparse *curp) { - struct man *man; - struct mdoc *mdoc; - struct roff *roff; - pdesc(curp); + /* + * 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. + */ - man = curp->man; - mdoc = curp->mdoc; - roff = curp->roff; + /* Zero the parse type. */ - if (MANDOCLEVEL_FATAL <= exit_status) + curp->mdoc = NULL; + curp->man = NULL; + file_status = MANDOCLEVEL_OK; + + /* 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 <= file_status) goto cleanup; /* NOTE a parser may not have been assigned, yet. */ - if ( ! (man || mdoc)) { + if ( ! (curp->man || curp->mdoc)) { fprintf(stderr, "%s: Not a manual\n", curp->file); - exit_status = MANDOCLEVEL_FATAL; + file_status = MANDOCLEVEL_FATAL; goto cleanup; } /* Clean up the parse routine ASTs. */ - if (mdoc && ! mdoc_endparse(mdoc)) { - assert(MANDOCLEVEL_FATAL <= exit_status); + if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) { + assert(MANDOCLEVEL_FATAL <= file_status); goto cleanup; } - if (man && ! man_endparse(man)) { - assert(MANDOCLEVEL_FATAL <= exit_status); + + if (curp->man && ! man_endparse(curp->man)) { + assert(MANDOCLEVEL_FATAL <= file_status); goto cleanup; } - if (roff && ! roff_endparse(roff)) { - assert(MANDOCLEVEL_FATAL <= exit_status); + + assert(curp->roff); + if ( ! roff_endparse(curp->roff)) { + assert(MANDOCLEVEL_FATAL <= file_status); goto cleanup; } @@ -480,7 +536,7 @@ fdesc(struct curparse *curp) * the requested level, do not produce output. */ - if (MANDOCLEVEL_OK != exit_status && curp->wstop) + if (MANDOCLEVEL_OK != file_status && curp->wstop) goto cleanup; /* If unset, allocate output dev now (if applicable). */ @@ -536,40 +592,53 @@ fdesc(struct curparse *curp) /* Execute the out device, if it exists. */ - if (man && curp->outman) - (*curp->outman)(curp->outdata, man); - if (mdoc && curp->outmdoc) - (*curp->outmdoc)(curp->outdata, mdoc); + 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)); - if (mdoc) - mdoc_reset(mdoc); - if (man) - man_reset(man); - if (roff) - roff_reset(roff); + /* 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; + /* + * 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)) { - exit_status = MANDOCLEVEL_SYSERR; + file_status = MANDOCLEVEL_SYSERR; return; } - if (NULL == curp->roff) - curp->roff = roff_alloc(&curp->regs, curp, mmsg); - assert(curp->roff); + /* Line number is per-file. */ curp->line = 1; + parsebuf(curp, blk, 1); if (with_mmap) @@ -582,26 +651,31 @@ static void parsebuf(struct curparse *curp, struct buf blk, int start) { struct buf ln; - int i, pos, lnn, of; + 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; - struct man *man; - struct mdoc *mdoc; - struct roff *roff; - man = curp->man; - mdoc = curp->mdoc; - roff = curp->roff; + /* + * 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; /* line number in the real file */ - pos = 0; /* byte number in the ln buffer */ + lnn = curp->line; + pos = 0; - for (i = 0; i < (int)blk.sz;) { + for (i = 0; i < (int)blk.sz; ) { if (0 == pos && '\0' == blk.buf[i]) break; - if (start) + + if (start) { curp->line = lnn; + curp->reparse_count = 0; + } while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) { if ('\n' == blk.buf[i]) { @@ -620,27 +694,33 @@ parsebuf(struct curparse *curp, struct buf blk, int st */ c = (unsigned char) blk.buf[i]; - if ( ! (isascii(c) && (isgraph(c) || isblank(c)))) { + + if ( ! (isascii(c) && + (isgraph(c) || isblank(c)))) { mmsg(MANDOCERR_BADCHAR, curp, curp->line, pos, "ignoring byte"); i++; continue; } - /* Trailing backslash is like a plain character. */ + /* 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 an escape and at least one other character. */ + + /* Found escape & at least one other char. */ + if ('\n' == blk.buf[i + 1]) { - /* Escaped newlines are skipped over */ i += 2; + /* Escaped newlines are skipped over */ ++lnn; continue; } + if ('"' == blk.buf[i + 1]) { i += 2; /* Comment, skip to end of line */ @@ -651,6 +731,7 @@ parsebuf(struct curparse *curp, struct buf blk, int st break; } } + /* Backout trailing whitespaces */ for (; pos > 0; --pos) { if (ln.buf[pos - 1] != ' ') @@ -660,7 +741,9 @@ parsebuf(struct curparse *curp, struct buf blk, int st } break; } - /* Some other escape sequence, copy and continue. */ + + /* Some other escape sequence, copy & cont. */ + if (pos + 1 >= (int)ln.sz) resize_buf(&ln, 256); @@ -670,6 +753,7 @@ parsebuf(struct curparse *curp, struct buf blk, int st if (pos >= (int)ln.sz) resize_buf(&ln, 256); + ln.buf[pos] = '\0'; /* @@ -682,11 +766,19 @@ parsebuf(struct curparse *curp, struct buf blk, int st */ of = 0; + rerun: - switch (roff_parseln(roff, curp->line, &ln.buf, &ln.sz, - of, &of)) { + rr = roff_parseln + (curp->roff, curp->line, + &ln.buf, &ln.sz, of, &of); + + switch (rr) { case (ROFF_REPARSE): - parsebuf(curp, ln, 0); + 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): @@ -698,7 +790,7 @@ rerun: pos = 0; continue; case (ROFF_ERR): - assert(MANDOCLEVEL_FATAL <= exit_status); + assert(MANDOCLEVEL_FATAL <= file_status); break; case (ROFF_SO): if (pfile(ln.buf + of, curp)) { @@ -717,35 +809,43 @@ rerun: * parsers with each one. */ - if ( ! (man || mdoc)) - pset(ln.buf + of, pos - of, curp, &man, &mdoc); + if ( ! (curp->man || curp->mdoc)) + pset(ln.buf + of, pos - of, curp); - /* Lastly, push down into the parsers themselves. */ + /* + * Lastly, push down into the parsers themselves. One + * of these will have already been set in the pset() + * routine. + */ - if (man && ! man_parseln(man, curp->line, ln.buf, of)) { - assert(MANDOCLEVEL_FATAL <= exit_status); - 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); + + if ( ! rc) { + assert(MANDOCLEVEL_FATAL <= file_status); + break; + } } - if (mdoc && ! mdoc_parseln(mdoc, curp->line, ln.buf, of)) { - assert(MANDOCLEVEL_FATAL <= exit_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, - struct man **man, struct mdoc **mdoc) +pset(const char *buf, int pos, struct curparse *curp) { int i; @@ -755,6 +855,9 @@ 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] || '\'' == buf[0]) { @@ -767,36 +870,38 @@ pset(const char *buf, int pos, struct curparse *curp, switch (curp->inttype) { case (INTT_MDOC): - if (NULL == curp->mdoc) - curp->mdoc = mdoc_alloc(&curp->regs, curp, mmsg); - assert(curp->mdoc); - *mdoc = curp->mdoc; + 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_alloc(&curp->regs, curp, mmsg); - assert(curp->man); - *man = curp->man; + 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_alloc(&curp->regs, curp, mmsg); - assert(curp->mdoc); - *mdoc = curp->mdoc; + 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_alloc(&curp->regs, curp, mmsg); - assert(curp->man); - *man = curp->man; + 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) { @@ -815,7 +920,6 @@ moptions(enum intt *tflags, char *arg) return(1); } - static int toptions(struct curparse *curp, char *arg) { @@ -844,7 +948,6 @@ toptions(struct curparse *curp, char *arg) return(1); } - static int woptions(struct curparse *curp, char *arg) { @@ -884,7 +987,6 @@ woptions(struct curparse *curp, char *arg) return(1); } - static int mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg) { @@ -906,8 +1008,8 @@ mmsg(enum mandocerr t, void *arg, int ln, int col, con fprintf(stderr, ": %s", msg); fputc('\n', stderr); - if (exit_status < level) - exit_status = level; + if (file_status < level) + file_status = level; return(level < MANDOCLEVEL_FATAL); }