Jim Blandy wrote:
> Jeff Johnston <jjohnstn@redhat.com> writes:
> 
>>Ok.  How about the following revised patch which modifies
>>typename_concat as suggested above plus incorporates your other
>>comments?
> 
> 
> Great --- thanks for revising this.
> 
> 
>>@@ -1599,7 +1601,11 @@ partial_die_full_name (struct partial_di
>>   if (parent_scope == NULL)
>>     return NULL;
>>   else
>>-    return concat (parent_scope, "::", pdi->name, NULL);
>>+    {
>>+      if (cu->language == language_java)
>>+	return concat (parent_scope, ".", pdi->name, NULL);
>>+      return concat (parent_scope, "::", pdi->name, NULL);
>>+    }
>> }
>> 
>> static void
> 
> 
> Any reason we're not using typename_concat here, too?
>

Done.

> 
>>@@ -3877,7 +3891,10 @@ read_namespace (struct die_info *die, st
>>       char *temp_name = alloca (strlen (previous_prefix)
>> 				+ 2 + strlen(name) + 1);
>>       strcpy (temp_name, previous_prefix);
>>-      strcat (temp_name, "::");
>>+      if (cu->language == language_java)
>>+	strcat (temp_name, ".");
>>+      else
>>+	strcat (temp_name, "::");
>>       strcat (temp_name, name);
>> 
>>       processing_current_prefix = temp_name;
> 
> 
> Here, too.  I guess you'll need to add a cleanup to free it, since
> typename_concat can't use alloca.
> 
> 

And done.

Ok now?

Retested on x86-linux with up to date compiler with needed gcj fixes.

-- Jeff J.