https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108489
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2023-01-23
Ever confirmed|0 |1
Status|UNCONFIRMED |NEW
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Hmm, I get a crash in
poly_int64 decl_bitsize;
if (mode == VOIDmode
|| (mode != BLKmode && ! direct_store[(int) mode]
&& GET_MODE_CLASS (mode) != MODE_COMPLEX_INT
&& GET_MODE_CLASS (mode) != MODE_COMPLEX_FLOAT)
|| REG_P (target)
|| GET_CODE (target) == SUBREG
/* If the field isn't aligned enough to store as an ordinary memref,
store it as a bit field. */
|| (mode != BLKmode
&& ((((MEM_ALIGN (target) < GET_MODE_ALIGNMENT (mode))
|| !multiple_p (bitpos, GET_MODE_ALIGNMENT (mode)))
&& targetm.slow_unaligned_access (mode, MEM_ALIGN (target)))
|| !multiple_p (bitpos, BITS_PER_UNIT)))
because we access MEM_ALIGN (target) on
(parallel:TI [
(expr_list:REG_DEP_TRUE (reg:SI 83 [ <retval> ])
(const_int 0 [0]))
(expr_list:REG_DEP_TRUE (reg:DI 84 [ <retval>+8 ])
(const_int 8 [0x8]))
])
looks like we don't expect a PARALLEL target here at all (the else expects a
MEM), then we should test that, or the caller needs to handle this. If
I simply adjust the guard with the following I get
diff --git a/gcc/expr.cc b/gcc/expr.cc
index 15be1c8db99..4ef11d0d8f7 100644
--- a/gcc/expr.cc
+++ b/gcc/expr.cc
@@ -7649,8 +7649,7 @@ store_field (rtx target, poly_int64 bitsize, poly_int64
bitpos,
|| (mode != BLKmode && ! direct_store[(int) mode]
&& GET_MODE_CLASS (mode) != MODE_COMPLEX_INT
&& GET_MODE_CLASS (mode) != MODE_COMPLEX_FLOAT)
- || REG_P (target)
- || GET_CODE (target) == SUBREG
+ || !MEM_P (target)
/* If the field isn't aligned enough to store as an ordinary memref,
store it as a bit field. */
|| (mode != BLKmode
t.C: In function 'ReturnObject example()':
t.C:24:1: error: unrecognizable insn:
24 | }
| ^
(insn 12 11 13 2 (set (reg:TI 86)
(parallel:TI [
(expr_list:REG_DEP_TRUE (reg:SI 83 [ <retval> ])
(const_int 0 [0]))
(expr_list:REG_DEP_TRUE (reg:DI 84 [ <retval>+8 ])
(const_int 8 [0x8]))
])) -1
(nil))
during RTL pass: vregs
so it seems the PARALLEL target needs more explicit handling.
