New bug in gas980329

New bug in gas980329

[Don Bowman]

--apologies for the earlier MIME/HTML. I'm having a fight with out
--corporate standard, MS Exchange. This message is just resent in
--plain text.

With the gas980329 drop, configured as mips-gnu-elf (from cygwin32),
the following fragment causes an assertion failure.

Interestingly, if I remove the .stabn, then it assembles OK.
I would have assumed the .stabn didn't affect the delays,
but I guess they must. Maybe to keep all of a 'C' line without
delay slots from the nex line?

------this fragment (f.s)-----
        lbu     $2,0($3)
LABEL:
        .stabn 68,0,4,LABEL
        addu    $6,$6,-1
-------------------------------

$ as -v -o f.o f.s
GNU assembler version 980329 (mips-gnu-elf), using BFD version 980329
f.s: Assembler messages:
f.s:3: Internal error!
Assertion failure in mips_emit_delays at ../../gas-980329/gas/config/tc-mips.c line 2280.
Please report this bug.


          for (l = insn_labels; l != NULL; l = l->next)
            {
              assert (S_GET_SEGMENT (l->label) == now_seg); <<<<<<<< This is line 2280
              l->label->sy_frag = frag_now;
              S_SET_VALUE (l->label, (valueT) frag_now_fix ());
              /* mips16 text labels are stored as odd.  */
              if (mips_opts.mips16)
                ++l->label->sy_value.X_add_number;
            }

Re: New bug in gas980329

[Ian Lance Taylor]

   From: Don Bowman <don@pixsci.com>
   Date: Mon, 30 Mar 1998 11:21:35 -0500

   With the gas980329 drop, configured as mips-gnu-elf (from cygwin32),
   the following fragment causes an assertion failure.

   Interestingly, if I remove the .stabn, then it assembles OK.
   I would have assumed the .stabn didn't affect the delays,
   but I guess they must. Maybe to keep all of a 'C' line without
   delay slots from the nex line?

   ------this fragment (f.s)-----
	   lbu     $2,0($3)
   LABEL:
	   .stabn 68,0,4,LABEL
	   addu    $6,$6,-1
   -------------------------------

This works fine with my sources.  I expect the problem is due to your
MIPS_STABS_ELF patch.

On the MIPS, a label is automatically attached to the following
statement.  You can see this happening in s_cons and s_align in
config/tc-mips.c.  Perhaps some of that code is getting executed
inappropriately, when there is some confusion about the section.
Perhaps you have switched to the .stab section without calling
mips_emit_delays.

Actually, you don't want to call mips_emit_delays in this case; you
just want to somehow prevent the section checks from firing while you
are in the .stab section.

Ian

New bug in gas980329

[Don Bowman]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1625 bytes --]

Title: New bug in gas980329





With the gas980329 drop, configured as mips-gnu-elf (from cygwin32),
the following fragment causes an assertion failure.


Interestingly, if I remove the .stabn, then it assembles OK.
I would have assumed the .stabn didn't affect the delays,
but I guess they must. Maybe to keep all of a 'C' line without
delay slots from the nex line?


------this fragment (f.s)-----
        lbu     $2,0($3)
LABEL:
        .stabn 68,0,4,LABEL
        addu    $6,$6,-1
-------------------------------


$ as -v -o f.o f.s
GNU assembler version 980329 (mips-gnu-elf), using BFD version 980329
f.s: Assembler messages:
f.s:3: Internal error!
Assertion failure in mips_emit_delays at ../../gas-980329/gas/config/tc-mips.c line 2280.
Please report this bug.



          for (l = insn_labels; l != NULL; l = l->next)
            {
              assert (S_GET_SEGMENT (l->label) == now_seg); <<<<<<<< This is line 2280
              l->label->sy_frag = frag_now;
              S_SET_VALUE (l->label, (valueT) frag_now_fix ());
              /* mips16 text labels are stored as odd.  */
              if (mips_opts.mips16)
                ++l->label->sy_value.X_add_number;
            }