vim: src/tag.c comparison

comparison src/tag.c @ 3131:52526aec4afb v7.3.336

updated for version 7.3.336 Problem: When a tags file specifies an encoding different from 'enc' it may hang and using a pattern doesn't work. Solution: Convert the whole line. Continue reading the header after the SORT tag. Add test83. (Yukihiro Nakadaira)

author	Bram Moolenaar <bram@vim.org>
date	Wed, 12 Oct 2011 19:53:52 +0200
parents	25672ad7f377
children	436b42f3cba1

comparison

equal deleted inserted replaced

-:91b43bbb3701
+:52526aec4afb
 					     other: minimal number of matches */
 char_u	*buf_ffname;		/* name of buffer for priority */
 {
 FILE       *fp;
 char_u     *lbuf;			/* line buffer */
+int		lbuf_size = LSIZE;	/* length of lbuf */
 char_u     *tag_fname;		/* name of tag file */
 tagname_T	tn;			/* info for get_tagfname() */
 int		first_file;		/* trying first tag file */
 tagptrs_T	tagp;
 int		did_open = FALSE;	/* did open a tag file */
 int		eof = FALSE;		/* found end-of-file */
 char_u	*p;
 char_u	*s;
 int		i;
 #ifdef FEAT_TAG_BINS
+int		tag_file_sorted = NUL;	/* !_TAG_FILE_SORTED value */
 struct tag_search_info	/* Binary search file offsets */
 {
 	off_t	low_offset;	/* offset for first char of first line that
 				   could match */
 	off_t	high_offset;	/* offset of char after last line that could
 char_u	*help_lang_find = NULL;		/* lang to be found */
 char_u	help_lang[3];			/* lang of current tags file */
 char_u	*saved_pat = NULL;		/* copy of pat[] */
 #endif
-/* Use two sets of variables for the pattern: "orgpat" holds the values
-* for the original pattern and "convpat" converted from 'encoding' to
-* encoding of the tags file.  "pats" point to either one of these. */
-pat_T	*pats;
 pat_T	orgpat;			/* holds unconverted pattern info */
 #ifdef FEAT_MBYTE
-pat_T	convpat;		/* holds converted pattern info */
 vimconv_T	vimconv;
 #endif
 #ifdef FEAT_TAG_BINS
 int		findall = (mincount == MAXCOL || mincount == TAG_MANY);
 #endif
 int		verbose = (flags & TAG_VERBOSE);
 help_save = curbuf->b_help;
 orgpat.pat = pat;
-pats = &orgpat;
 #ifdef FEAT_MBYTE
 vimconv.vc_type = CONV_NONE;
 #endif
 /*
 * Allocate memory for the buffers that are used
 */
-lbuf = alloc(LSIZE);
+lbuf = alloc(lbuf_size);
 tag_fname = alloc(MAXPATHL + 1);
 #ifdef FEAT_EMACS_TAGS
 ebuf = alloc(LSIZE);
 #endif
 for (mtt = 0; mtt < MT_COUNT; ++mtt)
 * Initialize a few variables
 */
 if (help_only)				/* want tags from help file */
 	curbuf->b_help = TRUE;			/* will be restored later */
-pats->len = (int)STRLEN(pat);
+orgpat.len = (int)STRLEN(pat);
 #ifdef FEAT_MULTI_LANG
 if (curbuf->b_help)
 {
 	/* When "@ab" is specified use only the "ab" language, otherwise
 	 * search all languages. */
-	if (pats->len > 3 && pat[pats->len - 3] == '@'
+	if (orgpat.len > 3 && pat[orgpat.len - 3] == '@'
-					  && ASCII_ISALPHA(pat[pats->len - 2])
+					  && ASCII_ISALPHA(pat[orgpat.len - 2])
-					 && ASCII_ISALPHA(pat[pats->len - 1]))
+					 && ASCII_ISALPHA(pat[orgpat.len - 1]))
 	{
-	    saved_pat = vim_strnsave(pat, pats->len - 3);
+	    saved_pat = vim_strnsave(pat, orgpat.len - 3);
 	    if (saved_pat != NULL)
 	    {
-		help_lang_find = &pat[pats->len - 2];
+		help_lang_find = &pat[orgpat.len - 2];
-		pats->pat = saved_pat;
+		orgpat.pat = saved_pat;
-		pats->len -= 3;
+		orgpat.len -= 3;
 	    }
 	}
 }
 #endif
-if (p_tl != 0 && pats->len > p_tl)		/* adjust for 'taglength' */
+if (p_tl != 0 && orgpat.len > p_tl)		/* adjust for 'taglength' */
-	pats->len = p_tl;
+	orgpat.len = p_tl;
-prepare_pats(pats, has_re);
+prepare_pats(&orgpat, has_re);
 #ifdef FEAT_TAG_BINS
 /* This is only to avoid a compiler warning for using search_info
 * uninitialised. */
 vim_memset(&search_info, 0, (size_t)1);
 * tags files twice.
 * When the tag file is case-fold sorted, it is either one or the other.
 * Only ignore case when TAG_NOIC not used or 'ignorecase' set.
 */
 #ifdef FEAT_TAG_BINS
-pats->regmatch.rm_ic = ((p_ic || !noic)
+orgpat.regmatch.rm_ic = ((p_ic || !noic)
-				&& (findall || pats->headlen == 0 || !p_tbs));
+				&& (findall || orgpat.headlen == 0 || !p_tbs));
 for (round = 1; round <= 2; ++round)
 {
-linear = (pats->headlen == 0 || !p_tbs || round == 2);
+linear = (orgpat.headlen == 0 || !p_tbs || round == 2);
 #else
-pats->regmatch.rm_ic = (p_ic || !noic);
+orgpat.regmatch.rm_ic = (p_ic || !noic);
 #endif
 /*
 * Try tag file names from tags option one by one.
 */
 #endif
 			break;			    /* end of file */
 		}
 	    }
 line_read_in:
+#ifdef FEAT_MBYTE
+	    if (vimconv.vc_type != CONV_NONE)
+	    {
+		char_u	*conv_line;
+		int	len;
+		/* Convert every line.  Converting the pattern from 'enc' to
+		 * the tags file encoding doesn't work, because characters are
+		 * not recognized. */
+		conv_line = string_convert(&vimconv, lbuf, NULL);
+		if (conv_line != NULL)
+		{
+		    /* Copy or swap lbuf and conv_line. */
+		    len = (int)STRLEN(conv_line) + 1;
+		    if (len > lbuf_size)
+		    {
+			vim_free(lbuf);
+			lbuf = conv_line;
+			lbuf_size = len;
+		    }
+		    else
+		    {
+			STRCPY(lbuf, conv_line);
+			vim_free(conv_line);
+		    }
+		}
+	    }
+#endif
 #ifdef FEAT_EMACS_TAGS
 	    /*
 	     * Emacs tags line with CTRL-L: New file name on next line.
 	     * The file name is followed by a ','.
 	     * When still at the start of the file, check for Emacs tags file
 	     * format, and for "not sorted" flag.
 	     */
 	    if (state == TS_START)
 	    {
+		if (STRNCMP(lbuf, "!_TAG_", 6) <= 0)
+		{
+		    /*
+		     * Read header line.
+		     */
+#ifdef FEAT_TAG_BINS
+		    if (STRNCMP(lbuf, "!_TAG_FILE_SORTED\t", 18) == 0)
+			tag_file_sorted = lbuf[18];
+#endif
+#ifdef FEAT_MBYTE
+		    if (STRNCMP(lbuf, "!_TAG_FILE_ENCODING\t", 20) == 0)
+		    {
+			/* Prepare to convert every line from the specified
+			 * encoding to 'encoding'. */
+			for (p = lbuf + 20; *p > ' ' && *p < 127; ++p)
+			    ;
+			*p = NUL;
+			convert_setup(&vimconv, lbuf + 20, p_enc);
+		    }
+#endif
+		    /* Read the next line.  Unrecognized flags are ignored. */
+		    continue;
+		}
+		/* Headers ends. */
 #ifdef FEAT_TAG_BINS
 		/*
 		 * When there is no tag head, or ignoring case, need to do a
 		 * linear search.
 		 * When no "!_TAG_" is found, default to binary search.  If
 		if (linear || use_cscope)
 # else
 		if (linear)
 # endif
 		    state = TS_LINEAR;
-		else if (STRNCMP(lbuf, "!_TAG_", 6) > 0)
+		else if (tag_file_sorted == NUL)
 		    state = TS_BINARY;
-		else if (STRNCMP(lbuf, "!_TAG_FILE_SORTED\t", 18) == 0)
+		else if (tag_file_sorted == '1')
-		{
-		    /* Check sorted flag */
-		    if (lbuf[18] == '1')
 			state = TS_BINARY;
-		    else if (lbuf[18] == '2')
+		else if (tag_file_sorted == '2')
-		    {
+		{
-			state = TS_BINARY;
+		    state = TS_BINARY;
-			sortic = TRUE;
+		    sortic = TRUE;
-			pats->regmatch.rm_ic = (p_ic || !noic);
+		    orgpat.regmatch.rm_ic = (p_ic || !noic);
-		    }
+		}
-		    else
+		else
-			state = TS_LINEAR;
+		    state = TS_LINEAR;
-		}
+		if (state == TS_BINARY && orgpat.regmatch.rm_ic && !sortic)
-		if (state == TS_BINARY && pats->regmatch.rm_ic && !sortic)
 		{
 		    /* binary search won't work for ignoring case, use linear
 		     * search. */
 		    linear = TRUE;
 		    state = TS_LINEAR;
 		    continue;
 		}
 #endif
 	    }
-#ifdef FEAT_MBYTE
-	    if (lbuf[0] == '!' && pats == &orgpat
-			   && STRNCMP(lbuf, "!_TAG_FILE_ENCODING\t", 20) == 0)
-	    {
-		/* Convert the search pattern from 'encoding' to the
-		 * specified encoding. */
-		for (p = lbuf + 20; *p > ' ' && *p < 127; ++p)
-		    ;
-		*p = NUL;
-		convert_setup(&vimconv, p_enc, lbuf + 20);
-		if (vimconv.vc_type != CONV_NONE)
-		{
-		    convpat.pat = string_convert(&vimconv, pats->pat, NULL);
-		    if (convpat.pat != NULL)
-		    {
-			pats = &convpat;
-			pats->len = (int)STRLEN(pats->pat);
-			prepare_pats(pats, has_re);
-			pats->regmatch.rm_ic = orgpat.regmatch.rm_ic;
-		    }
-		}
-		/* Prepare for converting a match the other way around. */
-		convert_setup(&vimconv, lbuf + 20, p_enc);
-		continue;
-	    }
-#endif
 	    /*
 	     * Figure out where the different strings are in this line.
 	     * For "normal" tags: Do a quick check if the tag matches.
 	     * This speeds up tag searching a lot!
 	     */
-	    if (pats->headlen
+	    if (orgpat.headlen
 #ifdef FEAT_EMACS_TAGS
 			    && !is_etag
 #endif
 					)
 	    {
 		 * there is no regexp, or the tag is too short.
 		 */
 		cmplen = (int)(tagp.tagname_end - tagp.tagname);
 		if (p_tl != 0 && cmplen > p_tl)	    /* adjust for 'taglength' */
 		    cmplen = p_tl;
-		if (has_re && pats->headlen < cmplen)
+		if (has_re && orgpat.headlen < cmplen)
-		    cmplen = pats->headlen;
+		    cmplen = orgpat.headlen;
-		else if (state == TS_LINEAR && pats->headlen != cmplen)
+		else if (state == TS_LINEAR && orgpat.headlen != cmplen)
 		    continue;
 #ifdef FEAT_TAG_BINS
 		if (state == TS_BINARY)
 		{
 		    /*
 		     * Compare the current tag with the searched tag.
 		     */
 		    if (sortic)
-			tagcmp = tag_strnicmp(tagp.tagname, pats->head,
+			tagcmp = tag_strnicmp(tagp.tagname, orgpat.head,
 							      (size_t)cmplen);
 		    else
-			tagcmp = STRNCMP(tagp.tagname, pats->head, cmplen);
+			tagcmp = STRNCMP(tagp.tagname, orgpat.head, cmplen);
 		    /*
 		     * A match with a shorter tag means to search forward.
 		     * A match with a longer tag means to search backward.
 		     */
 		    if (tagcmp == 0)
 		    {
-			if (cmplen < pats->headlen)
+			if (cmplen < orgpat.headlen)
 			    tagcmp = -1;
-			else if (cmplen > pats->headlen)
+			else if (cmplen > orgpat.headlen)
 			    tagcmp = 1;
 		    }
 		    if (tagcmp == 0)
 		    {
 		    /* No match yet and are at the end of the binary search. */
 		    break;
 		}
 		else if (state == TS_SKIP_BACK)
 		{
-		    if (MB_STRNICMP(tagp.tagname, pats->head, cmplen) != 0)
+		    if (MB_STRNICMP(tagp.tagname, orgpat.head, cmplen) != 0)
 			state = TS_STEP_FORWARD;
 		    else
 			/* Have to skip back more.  Restore the curr_offset
 			 * used, otherwise we get stuck at a long line. */
 			search_info.curr_offset = search_info.curr_offset_used;
 		    continue;
 		}
 		else if (state == TS_STEP_FORWARD)
 		{
-		    if (MB_STRNICMP(tagp.tagname, pats->head, cmplen) != 0)
+		    if (MB_STRNICMP(tagp.tagname, orgpat.head, cmplen) != 0)
 		    {
 			if ((off_t)ftell(fp) > search_info.match_offset)
 			    break;	/* past last match */
 			else
 			    continue;	/* before first match */
 		    }
 		}
 		else
 #endif
 		    /* skip this match if it can't match */
-		    if (MB_STRNICMP(tagp.tagname, pats->head, cmplen) != 0)
+		    if (MB_STRNICMP(tagp.tagname, orgpat.head, cmplen) != 0)
 		    continue;
 		/*
 		 * Can be a matching tag, isolate the file name and command.
 		 */
 	     */
 	    cmplen = (int)(tagp.tagname_end - tagp.tagname);
 	    if (p_tl != 0 && cmplen > p_tl)	    /* adjust for 'taglength' */
 		cmplen = p_tl;
 	    /* if tag length does not match, don't try comparing */
-	    if (pats->len != cmplen)
+	    if (orgpat.len != cmplen)
 		match = FALSE;
 	    else
 	    {
-		if (pats->regmatch.rm_ic)
+		if (orgpat.regmatch.rm_ic)
 		{
-		    match = (MB_STRNICMP(tagp.tagname, pats->pat, cmplen) == 0);
+		    match = (MB_STRNICMP(tagp.tagname, orgpat.pat, cmplen) == 0);
 		    if (match)
-			match_no_ic = (STRNCMP(tagp.tagname, pats->pat,
+			match_no_ic = (STRNCMP(tagp.tagname, orgpat.pat,
 								cmplen) == 0);
 		}
 		else
-		    match = (STRNCMP(tagp.tagname, pats->pat, cmplen) == 0);
+		    match = (STRNCMP(tagp.tagname, orgpat.pat, cmplen) == 0);
 	    }
 	    /*
 	     * Has a regexp: Also find tags matching regexp.
 	     */
 	    match_re = FALSE;
-	    if (!match && pats->regmatch.regprog != NULL)
+	    if (!match && orgpat.regmatch.regprog != NULL)
 	    {
 		int	cc;
 		cc = *tagp.tagname_end;
 		*tagp.tagname_end = NUL;
-		match = vim_regexec(&pats->regmatch, tagp.tagname, (colnr_T)0);
+		match = vim_regexec(&orgpat.regmatch, tagp.tagname, (colnr_T)0);
 		if (match)
 		{
-		    matchoff = (int)(pats->regmatch.startp[0] - tagp.tagname);
+		    matchoff = (int)(orgpat.regmatch.startp[0] - tagp.tagname);
-		    if (pats->regmatch.rm_ic)
+		    if (orgpat.regmatch.rm_ic)
 		    {
-			pats->regmatch.rm_ic = FALSE;
+			orgpat.regmatch.rm_ic = FALSE;
-			match_no_ic = vim_regexec(&pats->regmatch, tagp.tagname,
+			match_no_ic = vim_regexec(&orgpat.regmatch, tagp.tagname,
 								  (colnr_T)0);
-			pats->regmatch.rm_ic = TRUE;
+			orgpat.regmatch.rm_ic = TRUE;
 		    }
 		}
 		*tagp.tagname_end = cc;
 		match_re = TRUE;
 	    }
 			if (is_current)
 			    mtt = MT_GL_CUR;
 			else
 			    mtt = MT_GL_OTH;
 		    }
-		    if (pats->regmatch.rm_ic && !match_no_ic)
+		    if (orgpat.regmatch.rm_ic && !match_no_ic)
 			mtt += MT_IC_OFF;
 		    if (match_re)
 			mtt += MT_RE_OFF;
 		}
 		 * Store the info we need later, which depends on the kind of
 		 * tags we are dealing with.
 		 */
 		if (ga_grow(&ga_match[mtt], 1) == OK)
 		{
-#ifdef FEAT_MBYTE
-		    char_u	*conv_line = NULL;
-		    char_u	*lbuf_line = lbuf;
-		    if (vimconv.vc_type != CONV_NONE)
-		    {
-			/* Convert the tag line from the encoding of the tags
-			 * file to 'encoding'.  Then parse the line again. */
-			conv_line = string_convert(&vimconv, lbuf, NULL);
-			if (conv_line != NULL)
-			{
-			    if (parse_tag_line(conv_line,
-#ifdef FEAT_EMACS_TAGS
-					is_etag,
-#endif
-					&tagp) == OK)
-				lbuf_line = conv_line;
-			    else
-				/* doesn't work, go back to unconverted line. */
-				(void)parse_tag_line(lbuf,
-#ifdef FEAT_EMACS_TAGS
-						     is_etag,
-#endif
-						     &tagp);
-			}
-		    }
-#else
-# define lbuf_line lbuf
-#endif
 		    if (help_only)
 		    {
 #ifdef FEAT_MULTI_LANG
 # define ML_EXTRA 3
 #else
 			 * Emacs tag: <mtt><tag_fname><NUL><ebuf><NUL><lbuf>
 			 * other tag: <mtt><tag_fname><NUL><NUL><lbuf>
 			 * without Emacs tags: <mtt><tag_fname><NUL><lbuf>
 			 */
 			len = (int)STRLEN(tag_fname)
-						 + (int)STRLEN(lbuf_line) + 3;
+						 + (int)STRLEN(lbuf) + 3;
 #ifdef FEAT_EMACS_TAGS
 			if (is_etag)
 			    len += (int)STRLEN(ebuf) + 1;
 			else
 			    ++len;
 				s += STRLEN(ebuf) + 1;
 			    }
 			    else
 				*s++ = NUL;
 #endif
-			    STRCPY(s, lbuf_line);
+			    STRCPY(s, lbuf);
 			}
 		    }
 		    if (mfp != NULL)
 		    {
 			    ++match_count;
 			}
 			else
 			    vim_free(mfp);
 		    }
-#ifdef FEAT_MBYTE
-		    /* Note: this makes the values in "tagp" invalid! */
-		    vim_free(conv_line);
-#endif
 		}
 		else    /* Out of memory! Just forget about the rest. */
 		{
 		    retval = OK;
 		    stop_searching = TRUE;
 	    fclose(incstack[incstack_idx].fp);
 	    vim_free(incstack[incstack_idx].etag_fname);
 	}
 #endif
 #ifdef FEAT_MBYTE
-	if (pats == &convpat)
-	{
-	    /* Go back from converted pattern to original pattern. */
-	    vim_free(pats->pat);
-	    vim_free(pats->regmatch.regprog);
-	    orgpat.regmatch.rm_ic = pats->regmatch.rm_ic;
-	    pats = &orgpat;
-	}
 	if (vimconv.vc_type != CONV_NONE)
 	    convert_setup(&vimconv, NULL, NULL);
 #endif
 #ifdef FEAT_TAG_BINS
+	tag_file_sorted = NUL;
 	if (sort_error)
 	{
 	    EMSG2(_("E432: Tags file not sorted: %s"), tag_fname);
 	    sort_error = FALSE;
 	}
 	    tagname_free(&tn);
 #ifdef FEAT_TAG_BINS
 /* stop searching when already did a linear search, or when TAG_NOIC
 * used, and 'ignorecase' not set or already did case-ignore search */
-if (stop_searching || linear || (!p_ic && noic) || pats->regmatch.rm_ic)
+if (stop_searching || linear || (!p_ic && noic) || orgpat.regmatch.rm_ic)
 	  break;
 # ifdef FEAT_CSCOPE
 if (use_cscope)
 	  break;
 # endif
-pats->regmatch.rm_ic = TRUE;	/* try another time while ignoring case */
+orgpat.regmatch.rm_ic = TRUE;	/* try another time while ignoring case */
 }
 #endif
 if (!stop_searching)
 {
 	retval = OK;		/* It's OK even when no tag found */
 }
 findtag_end:
 vim_free(lbuf);
-vim_free(pats->regmatch.regprog);
+vim_free(orgpat.regmatch.regprog);
 vim_free(tag_fname);
 #ifdef FEAT_EMACS_TAGS
 vim_free(ebuf);
 #endif

Mercurial > vim

comparison src/tag.c @ 3131:52526aec4afb v7.3.336