Re: GCC build failed for native with your patch on 2003-04-11T21:30:01Z.

[Geoff Keating <geoffk at geoffk dot org>]

"GCC regression checker" <geoffk at apple dot com> writes:

> /Users/regress/tbox/native/build/gcc/xgcc -B/Users/regress/tbox/native/build/gcc/ -B/Users/regress/tbox/objs/powerpc-apple-darwin6.3/bin/ -B/Users/regress/tbox/objs/powerpc-apple-darwin6.3/lib/ -isystem /Users/regress/tbox/objs/powerpc-apple-darwin6.3/include -isystem /Users/regress/tbox/objs/powerpc-apple-darwin6.3/sys-include -O2  -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -isystem ./include   -g  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -I. -I. -I/Users/regress/tbox/cvs-gcc/gcc/gcc -I/Users/regress/tbox/cvs-gcc/gcc/gcc/. -I/Users/regress/tbox/cvs-gcc/gcc/gcc/config -I/Users/regress/tbox/cvs-gcc/gcc/gcc/../include -fexceptions -c /Users/regress/tbox/cvs-gcc/gcc/gcc/unwind-dw2.c -o libgcc/./unwind-dw2.o
> /Users/regress/tbox/cvs-gcc/gcc/gcc/unwind-dw2.c: In function `_Unwind_GetCFA':
> /Users/regress/tbox/cvs-gcc/gcc/gcc/unwind-dw2.c:181: warning: return makes integer from pointer without a cast
> [address=54001042 pc=00189cdc]
> In file included from /Users/regress/tbox/cvs-gcc/gcc/gcc/unwind-dw2.c:1265:
> /Users/regress/tbox/cvs-gcc/gcc/gcc/unwind.inc: In function `_Unwind_ForcedUnwind':
> /Users/regress/tbox/cvs-gcc/gcc/gcc/unwind.inc:207: internal compiler error: Segmentation Fault
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> make[3]: *** [libgcc/./unwind-dw2.o] Error 1
> make[2]: *** [stmp-multilib] Error 2
> make[1]: *** [stage1_build] Error 2
> make: *** [bootstrap] Error 2
> + '[' -s gcc/.bad_compare ']'
> + exit 1

This is Richard's patch:

+2003-04-11  Richard Henderson  <rth at redhat dot com>
+
+	* rtl.def (SYMBOL_REF): Add two 0 fields.
+	* gengtype.c (adjust_field_rtx_def): Handle them.
+	* print-rtl.c (print_rtx): Print them.
+	* rtl.h (SYMBOL_REF_DECL, SYMBOL_REF_FLAGS): New.
+	(SYMBOL_FLAG_FUNCTION, SYMBOL_REF_FUNCTION_P): New.
+	(SYMBOL_FLAG_LOCAL, SYMBOL_REF_LOCAL_P): New.
+	(SYMBOL_FLAG_SMALL, SYMBOL_REF_SMALL_P): New.
+	(SYMBOL_FLAG_TLS_SHIFT, SYMBOL_REF_TLS_MODEL): New.
+	(SYMBOL_FLAG_EXTERNAL, SYMBOL_REF_EXTERNAL_P): New.
+	(SYMBOL_FLAG_MACH_DEP): New.
+	* optabs.c (init_one_libfunc): Zap fake SYMBOL_REF_DECL.
+	* varasm.c (make_decl_rtl): Set SYMBOL_REF_DECL.
+	(assemble_static_space): Set SYMBOL_REF_FLAGS.
+	(assemble_trampoline_template): Likewise.
+	(output_constant_def, force_const_mem): Likewise.
+	(default_encode_section_info): New.
+	* output.h: Declare it.
+	* target-def.h (TARGET_ENCODE_SECTION_INFO): Use it.
+

The crash happens during a GC run.  In this SYMBOL_REF,

$1 = {
  code = SYMBOL_REF,
  mode = SImode,
  jump = 0,
  call = 0,
  unchanging = 0,
  volatil = 0,
  in_struct = 0,
  used = 0,
  integrated = 0,
  frame_related = 0,
  fld = {{
      rtwint = 19177330,
      rtint = 19177330,
      rtuint = 19177330,
      rtstr = 0x1249f72 "&L_memcpy$stub",
      rtx = 0x1249f72,
      rtvec = 0x1249f72,
      rttype = 19177330,
      rt_addr_diff_vec_flags = {
        min_align = 1,
        base_after_vec = 0,
        min_after_vec = 0,
        max_after_vec = 1,
        min_after_base = 0,
        max_after_base = 0,
        offset_unsigned = 1,
        0,
        scale = 159
      },
      rt_cselib = 0x1249f72,
      rtbit = 0x1249f72,
      rttree = 0x1249f72,
      bb = 0x1249f72,
      rtmem = 0x1249f72,
      rtreg = 0x1249f72
    }}
}

the third element, which you can't see in this dump, is a pointer
to uninitialised GCed memory.

It looks like one of the places that SYMBOL_REFs are created, perhaps
in the Darwin backend, doesn't properly initialise this new field.

-- 
- Geoffrey Keating <geoffk at geoffk dot org>