Re: String.clone may segfault since rev 14817

Hi,

At Sat, 5 Jan 2008 13:46:59 +0900,
Sam Ruby wrote in [ruby-core:14785]:
> At the point of failure, the value of *ptr as set on line 949 is NULL. 
> Apparently, under some rare set of circumstances, a request is made to 
> resize a string which is neither embedded nor is yet to be allocated on 
> the heap.  If I understood enough about what those circumstances might 
> be, I could probably produce a smaller test case, but my initial 
> attempts to model the failing code have failed to produce a segfault.

Could you send the smaller test case?

Anyway, seems no need to realloc for shorter string.


Index: string.c
===================================================================
--- string.c	(revision 14888)
+++ string.c	(working copy)
@@ -949,5 +949,5 @@ rb_str_resize(VALUE str, long len)
 	    char *ptr = RSTRING(str)->as.heap.ptr;
 	    STR_SET_EMBED(str);
-	    MEMCPY(RSTRING(str)->as.ary, ptr, char, len);
+	    if (len > 0) MEMCPY(RSTRING(str)->as.ary, ptr, char, len);
 	    RSTRING(str)->as.ary[len] = '\0';
 	    STR_SET_EMBED_LEN(str, len);


-- 
Nobu Nakada