Re: [avr-gcc-list] tool chain bug

[Theodore A. Roth]

On Thu, 26 Sep 2002, Marek Michalkiewicz wrote:

:) > The avr-gcc generate the following sequence:
:) > _PC_ = 2
:) > ...
:) > ...
:) > ...
:) > brne _PC_-NNN
:) >
:) > Old GAS resolved `_PC_-NNN' as constant expression.
:) > May be something changed in expression handling code in GAS :( and new
:) > GAS resolve `_PC_-NNN' as expression which must be corrected at link
:) > stage. But, as I remember, I don't support linking of 7-bit PC
:) > relative relocations (BFD_RELOC_AVR_7_PCREL).
:)
:) I guess the bug could be worked around in GCC by replacing all uses of _PC_
:) with one time labels.  Not very difficult, except that there are a lot of
:) places to change, it is easy to make a mistake, and 3.3 is close...

I don't think we need to do that since `.-NNN' works correctly in new gas
and old (see my conversation with Alan I cc'd you on). Is _PC_ always used
for pc relative insn (brXX)? Any other cases?

How close is gcc-3.3? If it isn't fixed, then when binutils-2.14 is out
and people use it with gcc-3.3, there's gonna be a lot of noise on the
avr-gcc-list list. ;-)

I'll see what I can do.

Ted Roth

avr-gcc-list at http://avr1.org