This is the mail archive of the
gcc-regression@gcc.gnu.org
mailing list for the GCC project.
Re: GCC build failed for native with your patch on 2004-01-27T03:45:02Z.
- From: Jan Hubicka <jh at suse dot cz>
- To: Geoff Keating <geoffk at geoffk dot org>
- Cc: gcc-regression at gcc dot gnu dot org, "CC:djee"@redhat.com,fjahanian at apple dot com, ghazi at caip dot rutgers dot edu, ian at wasabisystems dot com,jh at suse dot cz, kazu at cs dot umass dot edu, law at redhat dot com, mark at codesourcery dot com,pcarlini at suse dot de, ro at TechFak dot Uni-Bielefeld dot DE, rth at redhat dot com
- Date: Tue, 27 Jan 2004 22:12:56 +0100
- Subject: Re: GCC build failed for native with your patch on 2004-01-27T03:45:02Z.
- References: <20040127040331.928DBBD51DB4@gcc-regress.apple.com> <jmfze1kv68.fsf@desire.geoffk.org>
>
> > The last time the regression tests were run, GCC did not compile either,
> > so the problem might not be caused by your patch.
>
> The tester was a little confused about dates. The build failure was
> actually introduced between
>
> 2004-01-26T14:22:28Z
> and
> 2004-01-26T21:46:22Z
>
> The relevant changes appear to be:
>
> +2004-01-26 Jan Hubicka <jh@suse.cz>
> +
> + * cselib.c (discard_useless_values): Clear out value pointer pointing
> + to datastructure to be recycled.
> +
> +2004-01-25 Jan Hubicka <jh@suse.cz>
> +
> + * genextract.c (main): Do not output the memset when not checking.
> +
>
> +2004-01-26 Fariborz Jahanian <fjahanian@apple.com>
> +
> + * config/rs6000/rs6000.c (rs6000_emit_move): split slow
> + unaligned load/store into smaller loads and stores.
> +
> +2004-01-26 Fariborz Jahanian <fjahanian@apple.com>
> +
> + * function.c (assign_parms): Do not assign
> + long long argument to memory in prologue if
> + is it loaded into register.
> +
> +2004-01-26 Fariborz Jahanian <fjahanian@apple.com>
> +
> + PR middle-end/13779
> + * expr.c (emit_group_load): split constant
> + correctly into register components of PARALLEL insn.
> +
> +2004-01-26 Fariborz Jahanian <fjahanian@apple.com>
> +
> + * gcc/config/rs6000/rs6000.md (save_stack_nonlocal):
> + Use adjust_address_nv directly with appropriate mode.
> + (restore_stack_nonlocal): Ditto.
> +
>
> I suspect Jan's genextract change. The pattern that should be
> matching is:
>
> (define_insn ""
> [(set (match_operand:CCEQ 0 "cc_reg_operand" "=y,?y")
> (compare:CCEQ (match_operator:SI 1 "boolean_operator"
> [(match_operator:SI 2
> "branch_positive_comparison_operator"
> [(match_operand 3
> "cc_reg_operand" "y,y")
> (const_int 0)])
> (match_operator:SI 4
> "branch_positive_comparison_operator"
> [(match_operand 5
> "cc_reg_operand" "0,y")
> (const_int 0)])])
> (const_int 1)))]
> ""
> "cr%q1 %E0,%j2,%j4"
> [(set_attr "type" "cr_logical,delayed_cr")])
This looks somewhat strange, as the pattern must be refused by recog
that is called several times without clearing the arrays and definitly
shall not depend on it. Can I have testcase?
Honza
>
> --
> - Geoffrey Keating <geoffk@geoffk.org>