[PATCH] Fix PR37327, yet another problem with VRP and TREE_OVERFLOW
Richard Guenther
rguenther@suse.de
Tue Sep 2 13:54:00 GMT 2008
What a mess. The following hammer fixes this one.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to the trunk.
Richard.
2008-09-02 Richard Guenther <rguenther@suse.de>
PR tree-optimization/37327
* tree-vrp.c (register_new_assert_for): Make sure to not have
TREE_OVERFLOW set on the bound.
* gcc.c-torture/compile/pr37327.c: New testcase.
Index: gcc/tree-vrp.c
===================================================================
*** gcc/tree-vrp.c (revision 139886)
--- gcc/tree-vrp.c (working copy)
*************** register_new_assert_for (tree name, tree
*** 3785,3790 ****
--- 3785,3798 ----
&& gimple_code (gsi_stmt (si)) != GIMPLE_SWITCH);
#endif
+ /* Never build an assert comparing against an integer constant with
+ TREE_OVERFLOW set. This confuses our undefined overflow warning
+ machinery. */
+ if (TREE_CODE (val) == INTEGER_CST
+ && TREE_OVERFLOW (val))
+ val = build_int_cst_wide (TREE_TYPE (val),
+ TREE_INT_CST_LOW (val), TREE_INT_CST_HIGH (val));
+
/* The new assertion A will be inserted at BB or E. We need to
determine if the new location is dominated by a previously
registered location for A. If we are doing an edge insertion,
Index: gcc/testsuite/gcc.c-torture/compile/pr37327.c
===================================================================
*** gcc/testsuite/gcc.c-torture/compile/pr37327.c (revision 0)
--- gcc/testsuite/gcc.c-torture/compile/pr37327.c (revision 0)
***************
*** 0 ****
--- 1,48 ----
+ typedef signed char int8_t;
+ typedef short int int16_t;
+ typedef int int32_t;
+ typedef unsigned short int uint16_t;
+ typedef unsigned int uint32_t;
+ static inline uint32_t
+ safe_add_int8_t_s_s (int8_t si1, int16_t si2)
+ {
+ if ((si1) && (si2) && (si1 > (1 - si2)) || (si1) && (si2 < 0)
+ && (si1 < (-128 - si2)))
+ return si1;
+ return si1 + si2;
+ }
+
+ uint32_t g_2;
+ uint32_t g_113;
+ uint32_t g_145;
+ int32_t
+ func_17 (int32_t p_18, uint32_t p_19, uint32_t p_21)
+ {
+ uint32_t l_23 = -1L;
+ return l_23;
+ }
+
+ uint32_t
+ func_26 (uint16_t p_27)
+ {
+ uint32_t l_424;
+ if (func_93 (func_59 (safe_add_int8_t_s_s (p_27, 1))),
+ func_124 (l_424, -7L, 1, g_145, 1, (safe_add_int8_t_s_s (1, 1)), 1), 1,
+ 1, 1)
+ func_117 (1, 1,
+ (safe_add_uint64_t_u_u
+ (1, (safe_add_int8_t_s_s (1, 0xCDF4BE7A1B7E4629LL)))), 1);
+ uint32_t l_210;
+ if (func_17
+ ((safe_add_int8_t_s_s (g_2, (0x6C79A83AL | func_17 (1, 1, 1)))),
+ 0x4C9FL, 1))
+ {
+ uint32_t l_212;
+ if (safe_mul_int32_t_s_s
+ ((1, 1, l_212, (1, (safe_add_int8_t_s_s (l_210, 1)), 1, 1)), 1))
+ if (func_59 (1, (safe_add_int8_t_s_s (g_113, 1))))
+ {
+ }
+ }
+ }
+
More information about the Gcc-patches
mailing list