This is the mail archive of the
mailing list for the binutils project.
Re: [PATCH] fix gas/gcc incompatibility regression on sparc (PR gas/13441)
David Miller writes:
> From: Mikael Pettersson <email@example.com>
> Date: Sun, 27 Nov 2011 21:25:15 +0100
> > @@ -507,7 +509,7 @@ md_parse_option (int c, char *arg)
> > if (!architecture_requested
> > || opcode_arch > max_architecture)
> > max_architecture = opcode_arch;
> > - hwcap_allowed |= sa->hwcap_allowed;
> > + hwcap_allowed = sa->hwcap_allowed;
> > architecture_requested = 1;
> > }
> > break;
> BTW, this patch also has a problem here. This is intentionally an
> "or". Your change means that you can't do things like "-AX -AY" to
> turn on multiple sets of options any longer.
I did wonder about that "or", thanks for clarifying.
A different (and shorter) patch that solves my problem is to instead
qualify the hwcap_allowed test in sparc_ip with "&& architecture_requested",
but I'm guessing from your other comments that the behaviour change when
no -A was given is intentional so I won't try to push it upstream.