This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug target/85173] ICE in in extract_insn, at recog.c:2304 for -fstack-clash-protection on ARM
- From: "ktkachov at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Tue, 03 Apr 2018 12:51:29 +0000
- Subject: [Bug target/85173] ICE in in extract_insn, at recog.c:2304 for -fstack-clash-protection on ARM
- Auto-submitted: auto-generated
- References: <bug-85173-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85173
ktkachov at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |law at gcc dot gnu.org
--- Comment #2 from ktkachov at gcc dot gnu.org ---
The unrecognizable insn is:
(insn 29 28 30 5 (set (mem/v:SI (plus:SI (reg/f:SI 13 sp)
(const_int 16380 [0x3ffc])) [0 S4 A32])
(unspec:SI [
(const_int 0 [0])
] UNSPEC_PROBE_STACK)) "stack-check-3.c":23 -1
(nil))
The memory address is not valid for a store.
That address is created explicitly in anti_adjust_stack_and_probe_stack_clash
in explow.c:
1988 emit_stack_probe (plus_constant (Pmode, stack_pointer_rtx,
1989 (probe_interval
1990 - GET_MODE_SIZE
(word_mode))));
emit_stack_probe then passes that address directly to the target probe_stack.
Should the code in emit_stack_probe do a memory_address_addr_space_p check and
force_operand if the address is not valid?