Re: [PATCH 29/31] PDP11: Use `const_double_zero' to express double zero constant

["Maciej W. Rozycki" <macro@linux-mips.org>]

On Tue, 15 Dec 2020, Martin Liška wrote:

> If I see correctly, starting from this revision I can't compile a cross
> compiler of x86_64-linux-gnu:
> 
> ../configure --target=pdp11-aout --disable-bootstrap --enable-languages=c,c++
> --disable-multilib  --enable-obsolete

 Thanks for the heads-up and the details of the configuration, and sorry 
for the breakage.

> $ make
> build/genemit ../../gcc/common.md ../../gcc/config/pdp11/pdp11.md \
>   insn-conditions.md > tmp-emit.c
> genemit: Internal error: abort in gen_exp, at genemit.c:202
> make: *** [Makefile:2427: s-emit] Error 1
> 
> $ Breakpoint 1, fancy_abort (file=0x435238 "../../gcc/genemit.c", line=202,
> func=0x435230 "gen_exp") at ../../gcc/errors.c:133
> 133	  internal_error ("abort in %s, at %s:%d", func, trim_filename (file),
> line);
> (gdb) bt
> #0  fancy_abort (file=0x435238 "../../gcc/genemit.c", line=202, func=0x435230
> "gen_exp") at ../../gcc/errors.c:133
> #1  0x0000000000402e4b in gen_exp (x=0x470de0, subroutine_type=DEFINE_INSN,
> used=0x4aa4b0 "", info=0x7fffffffdea0) at ../../gcc/genemit.c:202
> #2  0x0000000000402f68 in gen_exp (x=0x4d6c70, subroutine_type=DEFINE_INSN,
> used=0x4aa4b0 "", info=0x7fffffffdea0) at ../../gcc/genemit.c:227
> #3  0x0000000000402f68 in gen_exp (x=0x4d6c50, subroutine_type=DEFINE_INSN,
> used=0x4aa4b0 "", info=0x7fffffffdea0) at ../../gcc/genemit.c:227
> #4  0x000000000040307d in gen_exp (x=0x4aa490, subroutine_type=DEFINE_INSN,
> used=0x4aa4b0 "", info=0x7fffffffdea0) at ../../gcc/genemit.c:255
> #5  0x00000000004036c1 in gen_insn (info=0x7fffffffdea0) at
> ../../gcc/genemit.c:435
> #6  0x0000000000404930 in main (argc=3, argv=0x7fffffffdfc8) at
> ../../gcc/genemit.c:911

 I don't have the target environment available and consequently the target 
does not complete building for me:

The directory that should contain system headers does not exist:
  .../usr/sysroot/usr/include
make[2]: *** [Makefile:3218: stmp-fixinc] Error 1

or otherwise I would have at least tried to compile it before submitting 
the change.  The configuration does build far enough to trigger this issue 
though and I can confirm the change is indeed the culprit.

> Can you please take a look?

 I'm fairly sure this is due to the difference in TARGET_SUPPORTS_WIDE_INT
with the VAX backend vs the PDP-11 one.  I have an idea how this should be 
addressed and will be implementing it shortly.

 NB the backend fails `-Werror' compilation:

In file included from ./tm.h:18,
                 from .../gcc/backend.h:28,
                 from .../gcc/varasm.c:31:
.../gcc/varasm.c: In function 'void assemble_zeros(long unsigned int)':
.../gcc/config/pdp11/pdp11.h:622:20: error: format '%o' expects argument of type 'unsigned int', but argument 3 has type 'long unsigned int' [-Werror=format=]
  622 |     fprintf (FILE, "\t.blkb\t%o\n", (SIZE) & 0xffff);   \
      |                    ^~~~~~~~~~~~~~~  ~~~~~~~~~~~~~~~
      |                                            |
      |                                            long unsigned int

etc., so that has to be disabled.  This may be with 64-bit hosts only, I'm 
not sure, however the way varargs are handled means the issue will likely 
cause a broken compiler to be built if the warning is ignored, at least 
with some hosts.

 Paul, can you please look into it sometime?

  Maciej