RE: Lots of "make check" regressions

[Alexandre Petit-Bianco <apbianco at cygnus dot com>]

Boehm, Hans writes:

> I can't build the gcj front end anymore.  The problem is the
> following piece of code around line 6800 in parse.y.

I saw that when I reviewed and tested Bryce's patch.  I thought I
mentioned that to him but I might have forgotten to, sorry.

In order to test Bryce's code, I just moved the local
declaration/assignment up a few lines. I'm going to check this in so
it won't prevent people from working. Both the branch and the trunk
seem to be affected, so I'll check it in both places.

./A

Index: ChangeLog
===================================================================
RCS file: /cvs/gcc/gcc/gcc/java/ChangeLog,v
retrieving revision 1.647.2.7
diff -u -p -r1.647.2.7 ChangeLog
--- ChangeLog	2001/02/19 22:46:57	1.647.2.7
+++ ChangeLog	2001/02/20 20:00:22
@@ -1,3 +1,8 @@
+2001-02-20  Alexandre Petit-Bianco  <apbianco@redhat.com>
+
+	* parse.y (check_inner_class_access): Moved declaration of local
+	`enclosing_decl_type' to the right location.
+
 2001-02-19  Bryce McKinlay  <bryce@albatross.co.nz>
 
 	* parse.y (parser_check_super_interface): Don't call 
Index: parse.y
===================================================================
RCS file: /cvs/gcc/gcc/gcc/java/parse.y,v
retrieving revision 1.251.2.5
diff -u -p -r1.251.2.5 parse.y
--- parse.y	2001/02/19 22:46:58	1.251.2.5
+++ parse.y	2001/02/20 20:00:58
@@ -6789,6 +6789,7 @@ check_inner_class_access (decl, enclosin
      tree decl, enclosing_decl, cl;
 {
   const char *access;
+  tree enclosing_decl_type;
 
   /* We don't issue an error message when CL is null. CL can be null
      as a result of processing a JDEP crafted by source_start_java_method
@@ -6798,7 +6799,7 @@ check_inner_class_access (decl, enclosin
   if (!decl || !cl)
     return;
 
-  tree enclosing_decl_type = TREE_TYPE (enclosing_decl);
+  enclosing_decl_type = TREE_TYPE (enclosing_decl);
 
   if (CLASS_PRIVATE (decl))
     {