https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79062
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2017-01-12
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.
info.fmtstr = get_format_string (info.format, &info.fmtloc);
if (!info.fmtstr)
return;
fails for info.format == "%lx":
(gdb) p debug_tree (info.format)
<addr_expr 0x7ffff66a7820
type <pointer_type 0x7ffff66b7000
type <array_type 0x7ffff66aef18 type <integer_type 0x7ffff66ae3f0 char>
SI
size <integer_cst 0x7ffff68a2e28 constant 32>
unit size <integer_cst 0x7ffff68a2e40 constant 4>
align 8 symtab 0 alias set -1 structural equality domain
<integer_type 0x7ffff66aee70>
pointer_to_this <pointer_type 0x7ffff66b7000>>
unsigned DI
size <integer_cst 0x7ffff68a2be8 constant 64>
unit size <integer_cst 0x7ffff68a2c00 constant 8>
align 64 symtab 0 alias set -1 structural equality>
readonly constant
arg 0 <string_cst 0x7ffff66a7800 type <array_type 0x7ffff66aef18>
readonly constant static "%lx\000">
t2.c:7:25 start: t2.c:7:25 finish: t2.c:7:25>
it runs into
if (TYPE_MAIN_VARIANT (TREE_TYPE (TREE_TYPE (format))) != char_type_node)
{
/* Wide format string. */
return NULL;
}
other code (in fold-const.c) simply checks for
&& (GET_MODE_CLASS (TYPE_MODE (TREE_TYPE (TREE_TYPE (string))))
== MODE_INT)
&& (GET_MODE_SIZE (TYPE_MODE (TREE_TYPE (TREE_TYPE (string)))) == 1)
char_type_node is _not_ "merged" with LTO as we allow combining
-f[un]singed-char units and have
for (i = 0; i < itk_none; i++)
/* Skip itk_char. char_type_node is dependent on -f[un]signed-char. */
if (i != itk_char)
record_common_node (cache, integer_types[i]);
you could also look at hwat tree-ssa-strlen.c does (TYPE_MODE (type) ==
TYPE_MODE (char_type_node), etc.)
