Comment 71 for bug 1890435

r10-4054 failed with

during RTL pass: reload
/root/qemu-5.0/fpu/softfloat.c: In function ‘sf_canonicalize’:
/root/qemu-5.0/fpu/softfloat.c:670:1: internal compiler error: Segmentation fault
  670 | }
      | ^

0x4b7443 crash_signal
 ../../gcc/gcc/toplev.c:326
0x66b7c7 thumb2_legitimate_address_p
 ../../gcc/gcc/config/arm/arm.c:8242
0x66bc31 arm_legitimate_address_p(machine_mode, rtx_def*, bool)
 ../../gcc/gcc/config/arm/arm.c:8657
0x458667 memory_address_addr_space_p(machine_mode, rtx_def*, unsigned char)
 ../../gcc/gcc/recog.c:1340
0x85b013 nonimmediate_soft_df_operand_1
 ../../gcc/gcc/config/arm/predicates.md:514
0x85b013 nonimmediate_soft_df_operand(rtx_def*, machine_mode)
 ../../gcc/gcc/config/arm/predicates.md:530
0x85b013 nonimmediate_soft_df_operand(rtx_def*, machine_mode)
 ../../gcc/gcc/config/arm/predicates.md:518

I'm unsure how to deal with this. It is an ICE, and it happened throughout the RTL stage as before. But the signature seems to be a different one.

It could be "bad" with the actual issue we look for hidden behind this.
It could be "good" as well with the actual issue we look fixed but hidden behind this.
Or it could be "bad" as in, this is the same issue but with a different signature.

For the time being I'll handle it as bad, with some luck we get our old trace back further down the bisect. But @Doko please have a look at the trace above (maybe it is a known one) and advise.

Starting on r10-2027