=================================================================== RCS file: /cvs/mandoc/read.c,v retrieving revision 1.93 retrieving revision 1.98 diff -u -p -r1.93 -r1.98 --- mandoc/read.c 2014/10/25 01:03:52 1.93 +++ mandoc/read.c 2014/11/26 23:42:14 1.98 @@ -1,4 +1,4 @@ -/* $Id: read.c,v 1.93 2014/10/25 01:03:52 schwarze Exp $ */ +/* $Id: read.c,v 1.98 2014/11/26 23:42:14 schwarze Exp $ */ /* * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons * Copyright (c) 2010-2014 Ingo Schwarze @@ -51,6 +51,7 @@ struct mparse { struct man *man; /* man parser */ struct mdoc *mdoc; /* mdoc parser */ struct roff *roff; /* roff parser (!NULL) */ + const struct mchars *mchars; /* character table */ char *sodest; /* filename pointed to by .so */ const char *file; /* filename of current input file */ struct buf *primary; /* buffer currently being parsed */ @@ -63,11 +64,12 @@ struct mparse { int filenc; /* encoding of the current file */ int reparse_count; /* finite interp. stack */ int line; /* line number in the file */ + pid_t child; /* the gunzip(1) process */ }; static void choose_parser(struct mparse *); static void resize_buf(struct buf *, size_t); -static void mparse_buf_r(struct mparse *, struct buf, int); +static void mparse_buf_r(struct mparse *, struct buf, size_t, int); static int read_whole_file(struct mparse *, const char *, int, struct buf *, int *); static void mparse_end(struct mparse *); @@ -302,27 +304,29 @@ choose_parser(struct mparse *curp) } /* - * 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'). + * Main parse routine for a buffer. + * It assumes encoding and line numbering are already set up. + * It can recurse directly (for invocations of user-defined + * macros, inline equations, and input line traps) + * and indirectly (for .so file inclusion). */ static void -mparse_buf_r(struct mparse *curp, struct buf blk, int start) +mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start) { const struct tbl_span *span; struct buf ln; + size_t pos; /* byte number in the ln buffer */ enum rofferr rr; - int i, of, rc; - int pos; /* byte number in the ln buffer */ + int of, rc; int lnn; /* line number in the real file */ unsigned char c; - memset(&ln, 0, sizeof(struct buf)); + memset(&ln, 0, sizeof(ln)); lnn = curp->line; pos = 0; - for (i = blk.offs; i < (int)blk.sz; ) { + while (i < blk.sz) { if (0 == pos && '\0' == blk.buf[i]) break; @@ -332,13 +336,11 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int if (lnn < 3 && curp->filenc & MPARSE_UTF8 && - curp->filenc & MPARSE_LATIN1) { - blk.offs = i; - curp->filenc = preconv_cue(&blk); - } + curp->filenc & MPARSE_LATIN1) + curp->filenc = preconv_cue(&blk, i); } - while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) { + while (i < blk.sz && (start || blk.buf[i] != '\0')) { /* * When finding an unescaped newline character, @@ -346,7 +348,7 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int * Skip a preceding carriage return, if any. */ - if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz && + if ('\r' == blk.buf[i] && i + 1 < blk.sz && '\n' == blk.buf[i + 1]) ++i; if ('\n' == blk.buf[i]) { @@ -360,7 +362,7 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int * case of 11 bytes: "\\[u10ffff]\0" */ - if (pos + 11 > (int)ln.sz) + if (pos + 11 > ln.sz) resize_buf(&ln, 256); /* @@ -369,13 +371,8 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int c = blk.buf[i]; if (c & 0x80) { - blk.offs = i; - ln.offs = pos; - if (curp->filenc && preconv_encode( - &blk, &ln, &curp->filenc)) { - pos = ln.offs; - i = blk.offs; - } else { + if ( ! (curp->filenc && preconv_encode( + &blk, &i, &ln, &pos, &curp->filenc))) { mandoc_vmsg(MANDOCERR_BADCHAR, curp, curp->line, pos, "0x%x", c); @@ -399,7 +396,7 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int /* Trailing backslash = a plain char. */ - if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) { + if (blk.buf[i] != '\\' || i + 1 == blk.sz) { ln.buf[pos++] = blk.buf[i++]; continue; } @@ -411,7 +408,7 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int * skip that one as well. */ - if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz && + if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz && '\n' == blk.buf[i + 2]) ++i; if ('\n' == blk.buf[i + 1]) { @@ -423,7 +420,7 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) { i += 2; /* Comment, skip to end of line */ - for (; i < (int)blk.sz; ++i) { + for (; i < blk.sz; ++i) { if ('\n' == blk.buf[i]) { ++i; ++lnn; @@ -460,7 +457,7 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int ln.buf[pos++] = blk.buf[i++]; } - if (pos >= (int)ln.sz) + if (pos >= ln.sz) resize_buf(&ln, 256); ln.buf[pos] = '\0'; @@ -497,20 +494,19 @@ mparse_buf_r(struct mparse *curp, struct buf blk, int [curp->secondary->sz] = '\0'; } rerun: - rr = roff_parseln(curp->roff, curp->line, - &ln.buf, &ln.sz, of, &of); + rr = roff_parseln(curp->roff, curp->line, &ln, &of); switch (rr) { case ROFF_REPARSE: if (REPARSE_LIMIT >= ++curp->reparse_count) - mparse_buf_r(curp, ln, 0); + mparse_buf_r(curp, ln, of, 0); else mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, pos, NULL); pos = 0; continue; case ROFF_APPEND: - pos = (int)strlen(ln.buf); + pos = strlen(ln.buf); continue; case ROFF_RERUN: goto rerun; @@ -521,8 +517,8 @@ rerun: assert(MANDOCLEVEL_FATAL <= curp->file_status); break; case ROFF_SO: - if (0 == (MPARSE_SO & curp->options) && - (i >= (int)blk.sz || '\0' == blk.buf[i])) { + if ( ! (curp->options & MPARSE_SO) && + (i >= blk.sz || blk.buf[i] == '\0')) { curp->sodest = mandoc_strdup(ln.buf + of); free(ln.buf); return; @@ -649,7 +645,6 @@ read_whole_file(struct mparse *curp, const char *file, return(0); } *with_mmap = 1; - fb->offs = 0; fb->sz = (size_t)st.st_size; fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0); if (fb->buf != MAP_FAILED) @@ -681,7 +676,6 @@ read_whole_file(struct mparse *curp, const char *file, ssz = read(fd, fb->buf + (int)off, fb->sz - off); if (ssz == 0) { fb->sz = off; - fb->offs = 0; return(1); } if (ssz == -1) { @@ -738,6 +732,7 @@ mparse_parse_buffer(struct mparse *curp, struct buf bl { struct buf *svprimary; const char *svfile; + size_t offset; static int recursion_depth; if (64 < recursion_depth) { @@ -758,11 +753,12 @@ mparse_parse_buffer(struct mparse *curp, struct buf bl (unsigned char)blk.buf[0] == 0xef && (unsigned char)blk.buf[1] == 0xbb && (unsigned char)blk.buf[2] == 0xbf) { - blk.offs = 3; + offset = 3; curp->filenc &= ~MPARSE_LATIN1; - } + } else + offset = 0; - mparse_buf_r(curp, blk, 1); + mparse_buf_r(curp, blk, offset, 1); if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status) mparse_end(curp); @@ -779,35 +775,31 @@ mparse_readmem(struct mparse *curp, const void *buf, s blk.buf = UNCONST(buf); blk.sz = len; - blk.offs = 0; mparse_parse_buffer(curp, blk, file); return(curp->file_status); } +/* + * If a file descriptor is given, use it and assume it points + * to the named file. Otherwise, open the named file. + * Read the whole file into memory and call the parsers. + * Called recursively when an .so request is encountered. + */ enum mandoclevel mparse_readfd(struct mparse *curp, int fd, const char *file) { struct buf blk; int with_mmap; int save_filenc; + pid_t save_child; - if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) { - curp->file_status = MANDOCLEVEL_SYSERR; - if (curp->mmsg) - (*curp->mmsg)(MANDOCERR_SYSOPEN, - curp->file_status, - file, 0, 0, strerror(errno)); - return(curp->file_status); - } + save_child = curp->child; + if (fd != -1) + curp->child = 0; + else if (mparse_open(curp, &fd, file) >= MANDOCLEVEL_SYSERR) + goto out; - /* - * 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, file, fd, &blk, &with_mmap)) { save_filenc = curp->filenc; curp->filenc = curp->options & @@ -822,38 +814,59 @@ mparse_readfd(struct mparse *curp, int fd, const char free(blk.buf); } - if (STDIN_FILENO != fd && -1 == close(fd)) + if (fd != STDIN_FILENO && close(fd) == -1) perror(file); + mparse_wait(curp); +out: + curp->child = save_child; return(curp->file_status); } enum mandoclevel -mparse_open(struct mparse *curp, int *fd, const char *file, - pid_t *child_pid) +mparse_open(struct mparse *curp, int *fd, const char *file) { int pfd[2]; + int save_errno; char *cp; enum mandocerr err; pfd[1] = -1; curp->file = file; + + /* Unless zipped, try to just open the file. */ + if ((cp = strrchr(file, '.')) == NULL || strcmp(cp + 1, "gz")) { - *child_pid = 0; - if ((*fd = open(file, O_RDONLY)) == -1) { - err = MANDOCERR_SYSOPEN; - goto out; - } - return(MANDOCLEVEL_OK); + curp->child = 0; + if ((*fd = open(file, O_RDONLY)) != -1) + return(MANDOCLEVEL_OK); + + /* Open failed; try to append ".gz". */ + + mandoc_asprintf(&cp, "%s.gz", file); + file = cp; + } else + cp = NULL; + + /* Before forking, make sure the file can be read. */ + + save_errno = errno; + if (access(file, R_OK) == -1) { + if (cp != NULL) + errno = save_errno; + err = MANDOCERR_SYSOPEN; + goto out; } + /* Run gunzip(1). */ + if (pipe(pfd) == -1) { err = MANDOCERR_SYSPIPE; goto out; } - switch (*child_pid = fork()) { + switch (curp->child = fork()) { case -1: err = MANDOCERR_SYSFORK; close(pfd[0]); @@ -876,11 +889,12 @@ mparse_open(struct mparse *curp, int *fd, const char * } out: + free(cp); *fd = -1; - *child_pid = 0; + curp->child = 0; curp->file_status = MANDOCLEVEL_SYSERR; if (curp->mmsg) - (*curp->mmsg)(err, curp->file_status, file, + (*curp->mmsg)(err, curp->file_status, curp->file, 0, 0, strerror(errno)); if (pfd[1] != -1) exit(1); @@ -888,11 +902,14 @@ out: } enum mandoclevel -mparse_wait(struct mparse *curp, pid_t child_pid) +mparse_wait(struct mparse *curp) { int status; - if (waitpid(child_pid, &status, 0) == -1) { + if (curp->child == 0) + return(MANDOCLEVEL_OK); + + if (waitpid(curp->child, &status, 0) == -1) { mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0, strerror(errno)); curp->file_status = MANDOCLEVEL_SYSERR; @@ -914,8 +931,8 @@ mparse_wait(struct mparse *curp, pid_t child_pid) } struct mparse * -mparse_alloc(int options, enum mandoclevel wlevel, - mandocmsg mmsg, const char *defos) +mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg, + const struct mchars *mchars, const char *defos) { struct mparse *curp; @@ -928,7 +945,8 @@ mparse_alloc(int options, enum mandoclevel wlevel, curp->mmsg = mmsg; curp->defos = defos; - curp->roff = roff_alloc(curp, options); + curp->mchars = mchars; + curp->roff = roff_alloc(curp, curp->mchars, options); if (curp->options & MPARSE_MDOC) curp->pmdoc = mdoc_alloc( curp->roff, curp, curp->defos,