[Bug rtl-optimization/64037] Miscompilation with LTO and enum class : char parameter

["hjl.tools at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64037

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ebotcazou at gcc dot gnu.org

--- Comment #10 from H.J. Lu <hjl.tools at gmail dot com> ---
To pass

enum class X : unsigned char {
  V = 2,
};

assign_parm_setup_reg calls ix86_promote_function_mode

(gdb) f 0
#0  ix86_promote_function_mode (type=0x7ffff19f85e8, mode=QImode, 
    punsignedp=0x7fffffffd3c4, fntype=0x7ffff19f8738, for_return=2)
    at /export/gnu/import/git/gcc/gcc/config/i386/i386.c:8310
8310      if (type != NULL_TREE && POINTER_TYPE_P (type))
(gdb) call debug_tree (type)
 <enumeral_type 0x7ffff19f85e8 X
    type <integer_type 0x7ffff18a93f0 unsigned char public unsigned string-flag
QI
        size <integer_cst 0x7ffff18a5fa8 constant 8>
        unit size <integer_cst 0x7ffff18a5fc0 constant 1>
        align 8 symtab 0 alias set -1 canonical type 0x7ffff18a93f0 precision 8
min <integer_cst 0x7ffff18a5fd8 0> max <integer_cst 0x7ffff18a5f78 255>>
    static unsigned type_5 QI size <integer_cst 0x7ffff18a5fa8 8> unit size
<integer_cst 0x7ffff18a5fc0 1>
    align 8 symtab 0 alias set -1 canonical type 0x7ffff19f85e8 precision 8 min
<integer_cst 0x7ffff18a5fd8 0> max <integer_cst 0x7ffff18a5f78 255>
    values <tree_list 0x7ffff19fb028
        purpose <identifier_node 0x7ffff19f6738 V
            bindings <(nil)>
            local bindings <(nil)>>
        value <const_decl 0x7ffff18c21c0 V type <enumeral_type 0x7ffff19f85e8
X>
            readonly constant used VOID file pr64037.ii line 2 col 3
            align 1 context <enumeral_type 0x7ffff19f85e8 X> initial
<integer_cst 0x7ffff19d8d08 2>>> context <translation_unit_decl 0x7ffff7ff91e0
D.1>
    chain <type_decl 0x7ffff19f5c78 X>>
(gdb) 

However, setup_incoming_promotions calls ix86_promote_function_mode with

Breakpoint 18, ix86_promote_function_mode (type=0x7ffff18a9690, mode=SImode, 
    punsignedp=0x7fffffffd774, fntype=0x7ffff19f8738, for_return=0)
    at /export/gnu/import/git/gcc/gcc/config/i386/i386.c:8310
8310      if (type != NULL_TREE && POINTER_TYPE_P (type))
(gdb) call debug_tree (type)
 <integer_type 0x7ffff18a9690 int public SI
    size <integer_cst 0x7ffff18a5e70 type <integer_type 0x7ffff18a9150
bitsizetype> constant 32>
    unit size <integer_cst 0x7ffff18a5e88 type <integer_type 0x7ffff18a90a8
sizetype> constant 4>
    align 32 symtab 0 alias set 1 canonical type 0x7ffff18a9690 precision 32
min <integer_cst 0x7ffff18c60c0 -2147483648> max <integer_cst 0x7ffff18c60d8
2147483647>
    pointer_to_this <pointer_type 0x7ffff18cb930>>
(gdb) 

It uses a different type and mode.

If I make this change to combine:

diff --git a/gcc/combine.c b/gcc/combine.c
index 1808f97..8bdfe2b 100644
--- a/gcc/combine.c
+++ b/gcc/combine.c
@@ -1562,7 +1562,7 @@ setup_incoming_promotions (rtx_insn *first)

       /* The mode and signedness of the argument as it is actually passed,
          after any TARGET_PROMOTE_FUNCTION_ARGS-driven ABI promotions.  */
-      mode3 = promote_function_mode (DECL_ARG_TYPE (arg), mode2, &uns3,
+      mode3 = promote_function_mode (DECL_ARG_TYPE (arg), mode1, &uns1,
                 TREE_TYPE (cfun->decl), 0);

to make setup_incoming_promotions consistent with assign_parm_setup_reg,
the testcase works.