| Message ID | 56293685.7040507@redhat.com |
|---|---|
| State | New |
| Headers | show |
On Thu, Oct 22, 2015 at 01:18:29PM -0600, Jeff Law wrote: > > Various ports are currently failing to build, faulting in > write_eligible_delay. > > This can happen if the target has delay slots defined, but does not have > annul-true or annul-false slots. cris is a reasonable example. > > This is most easily addressed by writing out a trivial dummy function if we > don't have annul-true or annul-false slots. > > I've verified this fixes the 47 newly failing targets in config-list.mk and > survives the usual bootstrap & comparison test on x86_64-linux-gnu. thanks for taking care of this. > +write_dummy_eligible_delay (FILE *outf, const char *kind) > +{ > + /* Write function prelude. */ > + > + fprintf (outf, "int\n"); > + fprintf (outf, "eligible_for_%s (rtx_insn *delay_insn ATTRIBUTE_UNUSED,\n" > + " int slot ATTRIBUTE_UNUSED,\n" > + " rtx_insn *candidate_insn ATTRIBUTE_UNUSED,\n" > + " int flags ATTRIBUTE_UNUSED)\n", > + kind); Couldn't you just leave the arguments unnamed as a simpler way to avoid unused warnings? Trev
On 10/22/2015 01:59 PM, Trevor Saunders wrote: >> >> I've verified this fixes the 47 newly failing targets in config-list.mk and >> survives the usual bootstrap & comparison test on x86_64-linux-gnu. > > thanks for taking care of this. No problem. > >> +write_dummy_eligible_delay (FILE *outf, const char *kind) >> +{ >> + /* Write function prelude. */ >> + >> + fprintf (outf, "int\n"); >> + fprintf (outf, "eligible_for_%s (rtx_insn *delay_insn ATTRIBUTE_UNUSED,\n" >> + " int slot ATTRIBUTE_UNUSED,\n" >> + " rtx_insn *candidate_insn ATTRIBUTE_UNUSED,\n" >> + " int flags ATTRIBUTE_UNUSED)\n", >> + kind); > > Couldn't you just leave the arguments unnamed as a simpler way to avoid > unused warnings? Had I known about that feature, I would have considered it! I know for the future though. jeff
diff --git a/gcc/ChangeLog b/gcc/ChangeLog index 08d26a9..04d8e16 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,9 @@ +2015-10-22 Jeff Law <law@redhat.com> + + * genattrtab.c (main): If we do not have any annul-true or annul-false + slots, then write out a dummy eligible_for_annul_true or + eligible_for_annul_false as needed. + 2015-10-22 Nick Clifton <nickc@redhat.com> * config/msp430/msp430.opt: Add -msilicon-errata and diff --git a/gcc/genattrtab.c b/gcc/genattrtab.c index 8d1fa6c..32b837c 100644 --- a/gcc/genattrtab.c +++ b/gcc/genattrtab.c @@ -4411,6 +4411,26 @@ write_indent (FILE *outf, int indent) fprintf (outf, " "); } +/* If the target does not have annul-true or annul-false delay slots, this + function will create a dummy eligible_for function on OUTF which always + returns false. KIND will be annul_true or annul_false. */ + +static void +write_dummy_eligible_delay (FILE *outf, const char *kind) +{ + /* Write function prelude. */ + + fprintf (outf, "int\n"); + fprintf (outf, "eligible_for_%s (rtx_insn *delay_insn ATTRIBUTE_UNUSED,\n" + " int slot ATTRIBUTE_UNUSED,\n" + " rtx_insn *candidate_insn ATTRIBUTE_UNUSED,\n" + " int flags ATTRIBUTE_UNUSED)\n", + kind); + fprintf (outf, "{\n"); + fprintf (outf, " return 0;\n"); + fprintf (outf, "}\n\n"); +} + /* Write a subroutine that is given an insn that requires a delay slot, a delay slot ordinal, and a candidate insn. It returns nonzero if the candidate can be placed in the specified delay slot of the insn. @@ -5307,8 +5327,14 @@ main (int argc, char **argv) (The function to compute the number of delay slots will be written below.) */ write_eligible_delay (attr_file, "delay"); - write_eligible_delay (attr_file, "annul_true"); - write_eligible_delay (attr_file, "annul_false"); + if (have_annul_true) + write_eligible_delay (attr_file, "annul_true"); + else + write_dummy_eligible_delay (attr_file, "annul_true"); + if (have_annul_false) + write_eligible_delay (attr_file, "annul_false"); + else + write_dummy_eligible_delay (attr_file, "annul_false"); /* Write out constant delay slot info. */ write_const_num_delay_slots (attr_file);
Various ports are currently failing to build, faulting in write_eligible_delay. This can happen if the target has delay slots defined, but does not have annul-true or annul-false slots. cris is a reasonable example. This is most easily addressed by writing out a trivial dummy function if we don't have annul-true or annul-false slots. I've verified this fixes the 47 newly failing targets in config-list.mk and survives the usual bootstrap & comparison test on x86_64-linux-gnu. Installed on the trunk. Jeff commit 6119ad4e875ff0c856102c14b08d869c7d33b357 Author: law <law@138bc75d-0d04-0410-961f-82ee72b054a4> Date: Thu Oct 22 19:18:05 2015 +0000 [PATCH] Fix abort in write_eligible_delay * genattrtab.c (main): If we do not have any annul-true or annul-false slots, then write out a dummy eligible_for_annul_true or eligible_for_annul_false as needed. git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@229184 138bc75d-0d04-0410-961f-82ee72b054a4