From: "K.Kosako" <kosako@...>
Date: 2002-11-29T17:24:58+09:00
Subject: [ruby-dev:18973] Re: [ONIGURUMA] \S doesn't match multibyte char

Yukihiro Matsumotoさんの
<1038550684.556438.10927.nullmailer@picachu.netlab.jp>から
> 1.7ではRSTRING(str)->ptrがNULLのことがあるので、以下のパッチ
> が必要でした。
> 
> --- regexec.c	24 Apr 2002 07:50:39 -0000	1.1.1.1
> +++ regexec.c	29 Nov 2002 06:17:13 -0000

どうもすいません。ruby-bugs-jaの
% ./ruby -e 'String.new.sub!(/\b/, "")'
のことですよね。

私としては、以下のように直したいと思います。
rangeの意味は、forward searchのときには探索右端 + 1で作っているので。


--- regexec.c.orig	Wed Apr 24 12:01:50 2002
+++ regexec.c	Fri Nov 29 17:10:55 2002
@@ -2235,8 +2235,8 @@ regex_search(regex_t* reg, UChar* str, U
   reg->state++;  /* increment as search counter */
 
 #ifdef REG_DEBUG_SEARCH
-  fprintf(stderr, "regex_search (entry point): end: %d, start: %d, range: %d\n",
-	  (int )(end - str), (int )(start - str), (int )(range - str));
+  fprintf(stderr, "regex_search (entry point): str: %d, end: %d, start: %d, range: %d\n",
+	  (int )str, (int )(end - str), (int )(start - str), (int )(range - str));
 #endif
 
   r = regex_region_resize(region, reg->max_mem + 1);
@@ -2244,6 +2244,13 @@ regex_search(regex_t* reg, UChar* str, U
 
   if (start > end || start < str) goto mismatch;
 
+#define MATCH_AND_RETURN_CHECK \
+  r = match_at(reg, str, end, s, prev, region, &msa);\
+  if (r != REG_MISMATCH) {\
+    if (r >= 0) goto match;\
+    goto finish; /* error */ \
+  }
+
   /* anchor optimize: resume search range */
   if (reg->anchor != 0 && str < end) {
     UChar* semi_end;
@@ -2320,19 +2327,24 @@ regex_search(regex_t* reg, UChar* str, U
       goto begin_position;
     }
   }
+  else if (str == end) { /* empty string */
+    static UChar* address_for_empty_string = "";
 
 #ifdef REG_DEBUG_SEARCH
-  fprintf(stderr, "regex_search(apply anchor): end: %d, start: %d, range: %d\n",
-	  (int )(end - str), (int )(start - str), (int )(range - str));
+    fprintf(stderr, "regex_search: empty string.\n");
 #endif
 
-#define MATCH_AND_RETURN_CHECK \
-  r = match_at(reg, str, end, s, prev, region, &msa);\
-  if (r != REG_MISMATCH) {\
-    if (r >= 0) goto match;\
-    goto finish; /* error */ \
+    s = start = end = str = address_for_empty_string;
+    prev = (UChar* )NULL;
+    MATCH_AND_RETURN_CHECK;
+    goto mismatch;
   }
 
+#ifdef REG_DEBUG_SEARCH
+  fprintf(stderr, "regex_search(apply anchor): end: %d, start: %d, range: %d\n",
+	  (int )(end - str), (int )(start - str), (int )(range - str));
+#endif
+
   s = start;
   if (range > start) {   /* forward search */
     if (s > str)
--
小迫＠ソフネック   渋谷区恵比寿1-15-1