This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] Fix TLS related ICE on IA-32
- From: Jakub Jelinek <jakub at redhat dot com>
- To: Richard Henderson <rth at redhat dot com>, gcc-patches at gcc dot gnu dot org, drepper at redhat dot com, roland at redhat dot com
- Date: Tue, 17 Sep 2002 02:54:11 -0400
- Subject: Re: [PATCH] Fix TLS related ICE on IA-32
- References: <20020911001002.L1013@sunsite.ms.mff.cuni.cz> <20020917043517.GO19591@redhat.com>
- Reply-to: Jakub Jelinek <jakub at redhat dot com>
On Mon, Sep 16, 2002 at 09:35:17PM -0700, Richard Henderson wrote:
> On Wed, Sep 11, 2002 at 12:10:02AM +0200, Jakub Jelinek wrote:
> > * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
> > UNSPEC_NTPOFF and UNSPEC_DTPOFF to be offsetted by constant.
> > * gcc.dg/tls/run-1.c: New test.
> > * gcc.dg/tls/run-2.c: New test.
> Did we come to a decision about "4+foo@ntpoff" vs "-4+foo@ntpoff"?
@ntpoff is not a problem, the problem was @tpoff. I followed your
and others request, so the counterpart of
addl $4+foo@ntpoff, %eax
subl $-4+foo@tpoff, %eax
subl $foo@tpoff, %eax
addl $4, %eax
But as we don't know what exactly Sun linker does with addends,
it is probably better to use the 2 insn sequence for -mtls-dialect=sun.