This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
hppa .reg
- To: binutils at sourceware dot cygnus dot com
- Subject: hppa .reg
- From: Alan Modra <alan at linuxcare dot com dot au>
- Date: Wed, 23 Aug 2000 20:41:02 +1000 (EST)
- cc: parisc-linux at thepuffingroup dot com, parisc at lists dot linuxcare dot com, Jeffrey A Law <law at cygnus dot com>
Hi,
I've just spent a bit of time modifying tc-hppa.c:pa_parse_number and
sundry places where it's called to allow:
pte: .reg %r16
depd pte,31,32,pte
Previously, gas would barf on lack of a `%' followed by one of the
predefined regs for any FLAG_STRICT insn like the above `depd'.
All well and good, but my patch puts these register alias symbols into
reg_section, the natural place for them to go, and this happens to kill
one of the hppa tests. gas/testsuite/gas/hppa/parse/regpopbug.s is like
this:
--------
.code
r0: .reg %r0
shift: .reg %sar
fpreg10: .reg %fr10
shift2: .reg shift
; Make sure we didn't botch .equ...
yabba: .equ r0 + shift
---------
and adding two register symbols ain't an absolute expression for the
`.equ', so I've "botched .equ"
Now the question is, does adding two reg symbols make sense (it doesn't
to me), or is this particular case a backwards compatibility issue? Can
someone tell me what the current hp assembler does with the above test?
Regards, Alan Modra
--
Linuxcare. Support for the Revolution.