Re: [avr-gcc-list] tool chain bug (fwd)

["Theodore A. Roth" <troth at verinet dot com>]

Alan,

Thanks for the help.

On Thu, 26 Sep 2002, Alan Modra wrote:

:)Heh, large changes get blamed for everything. :)  Looking over the
:)part I removed from tc-avr.c again though, I think I may have
:)removed too much.  Does the following patch cure your problem?

No. value is wrong coming into md_apply_fix3(). Consider this code snippet:

[troth@bozoland messy]$ cat mm.s
        .arch atmega128
__SREG__ = 0x3f
__SP_H__ = 0x3e
__SP_L__ = 0x3d
__tmp_reg__ = 0
__zero_reg__ = 1
_PC_ = 2
.Lfoo:
        dec r24
        brne _PC_-4
.Lbar:
        dec r24
        brne .Lbar
.Lloop:
        rjmp .Lloop

Running this through gas with your patch yields this:

00000000 <.text>:
   0:   8a 95           dec     r24
   2:   e9 f7           brne    .-6             ; 0xfffffffe
   4:   8a 95           dec     r24
   6:   f1 f7           brne    .-4             ; 0x4
   8:   ff cf           rjmp    .-2             ; 0x8

line "2:" should match line "6:".

Somehow valP is getting mangled before it gets passed to md_apply_fix3().

I tried adding this ugliness on top of your patch, but alas, it's still 
broken:

Index: gas/config/tc-avr.c
===================================================================
RCS file: /cvs/src/src/gas/config/tc-avr.c,v
retrieving revision 1.18
diff -u -r1.18 tc-avr.c
--- gas/config/tc-avr.c 5 Sep 2002 00:01:17 -0000   1.18
+++ gas/config/tc-avr.c 26 Sep 2002 03:38:32 -0000
@@ -705,8 +705,11 @@

     case 'l':
       str = parse_exp (str, &op_expr);
-      fix_new_exp (frag_now, where, opcode->insn_size * 2,
-          &op_expr, true, BFD_RELOC_AVR_7_PCREL);
+      if (op_expr.X_add_symbol)
+        fix_new_exp (frag_now, where, opcode->insn_size * 2,
+                     &op_expr, true, BFD_RELOC_AVR_7_PCREL);
+      else
+        op_mask |= (((op_expr.X_add_number) << 3) & 0x3f8);
       break;

     case 'i':

Ted Roth

:)
:)	* gas/config/tc-avr.c (md_apply_fix3): Reinstate code handling
:)	pcrel fixups to current or absolute section.
:)
:)Index: gas/config/tc-avr.c
:)===================================================================
:)RCS file: /cvs/src/src/gas/config/tc-avr.c,v
:)retrieving revision 1.18
:)diff -u -p -r1.18 tc-avr.c
:)--- gas/config/tc-avr.c	5 Sep 2002 00:01:17 -0000	1.18
:)+++ gas/config/tc-avr.c	26 Sep 2002 00:21:55 -0000
:)@@ -838,6 +838,17 @@ md_apply_fix3 (fixP, valP, seg)
:)   if (fixP->fx_addsy == (symbolS *) NULL)
:)     fixP->fx_done = 1;
:) 
:)+  else if (fixP->fx_pcrel)
:)+    {
:)+      segT s = S_GET_SEGMENT (fixP->fx_addsy);
:)+
:)+      if (s == seg || s == absolute_section)
:)+	{
:)+	  value += S_GET_VALUE (fixP->fx_addsy);
:)+	  fixP->fx_done = 1;
:)+	}
:)+    }
:)+
:)   /* We don't actually support subtracting a symbol.  */
:)   if (fixP->fx_subsy != (symbolS *) NULL)
:)     as_bad_where (fixP->fx_file, fixP->fx_line, _("expression too complex"));
:)
:)
:)> >From what Denis says and my time spent in the debugger, I think that this
:)> fragment from tc-avr.c:
:)> 
:)>     706     case 'l':
:)>     707       str = parse_exp (str, &op_expr);
:)>     708       fix_new_exp (frag_now, where, opcode->insn_size * 2,
:)>     709                    &op_expr, true, BFD_RELOC_AVR_7_PCREL);
:)>     710       break;
:)>     711
:)> 
:)> should probably not be using a fixup and instead just calculate the
:)> op_mask from the pcrel offset and be done with it.
:)
:)I agree.
:)
:)> stage. But, as I remember, I don't support linking of 7-bit PC
:)> relative relocations (BFD_RELOC_AVR_7_PCREL).
:)
:)Especially in light of Denis' comment.
:)
:)