This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: The COFF patch causes testsuit failure on COFF
- To: Nick Clifton <nickc at redhat dot com>
- Subject: Re: The COFF patch causes testsuit failure on COFF
- From: "H . J . Lu" <hjl at lucon dot org>
- Date: Tue, 27 Mar 2001 18:37:52 -0800
- Cc: nick at coelacanth dot com, binutils at sourceware dot cygnus dot com
- References: <20010327160630.A11673@lucon.org> <p066gueife.fsf@elmo.cygnus.com>
On Tue, Mar 27, 2001 at 06:28:05PM -0800, Nick Clifton wrote:
> Hi H.J, Hi Nick
>
> > The patch
> >
> > 2001-03-27 Nick Papadonis <nick@coelacanth.com>
> >
> > * read.c (equals): (for COFF) default symbols to being local.
> >
> > caused testsuite/gas/macros/test1.exp to fail on COFF. test1.s has
>
> The patch below fixes this regression.
>
> Unfortunately I have not yet been able to determine why the arm-coff
> port did not show this regression.
>
> Nick - please can you try this patch out with your m68k-coff toolchain
> and let me know if it reintroduces the bug that your patch was trying
> to fix.
>
I tried. It didn't work for the m68k-coff test. This one seems to work.
H.J.
----
Index: read.c
===================================================================
RCS file: /work/cvs/gnu/binutils/gas/read.c,v
retrieving revision 1.27
diff -u -p -r1.27 read.c
--- read.c 2001/03/27 23:56:21 1.27
+++ read.c 2001/03/28 02:34:47
@@ -4857,6 +4857,13 @@ equals (sym_name, reassign)
}
else
{
+#ifdef OBJ_COFF
+ int local;
+
+ symbolP = symbol_find (sym_name);
+ local = symbolP == NULL;
+ if (local)
+#endif /* OBJ_COFF */
symbolP = symbol_find_or_make (sym_name);
/* Permit register names to be redefined. */
if (!reassign
@@ -4866,7 +4873,8 @@ equals (sym_name, reassign)
#ifdef OBJ_COFF
/* "set" symbols are local unless otherwise specified. */
- SF_SET_LOCAL (symbolP);
+ if (local)
+ SF_SET_LOCAL (symbolP);
#endif /* OBJ_COFF */
pseudo_set (symbolP);