https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95757
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Blocks| |85316
CC| |amacleod at redhat dot com
Priority|P3 |P1
--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
So the warning machinery (or likely the range machinery) is confused about
size_t r_1_2 = UR (1, 2);
size_t r_2_3 = UR (2, 3);
T (char, S (2), r_1_2); // { dg-warning "\\\[-Wstringop-overflow"
"pr92814" }
T (char, S (0), r_2_3);
but it treats the case well when the second T is elided or when the size
computation is moved before the Ts like with
diff --git a/gcc/testsuite/gcc.dg/Wstringop-overflow-25.c
b/gcc/testsuite/gcc.dg/Wstringop-overflow-25.c
index bc609587575..116fbf600a6 100644
--- a/gcc/testsuite/gcc.dg/Wstringop-overflow-25.c
+++ b/gcc/testsuite/gcc.dg/Wstringop-overflow-25.c
@@ -367,16 +367,16 @@ struct Flex
NOIPA void test_strcpy_malloc_flexarray (void)
{
size_t r_0_1 = UR (0, 1);
- size_t r_1_2 = UR (1, 2);
- size_t r_2_3 = UR (2, 3);
T (char, S (0), r_0_1);
T (char, S (1), r_0_1); // { dg-warning "\\\[-Wstringop-overflow"
"pr92814" }
+ size_t r_1_2 = UR (1, 2);
T (char, S (0), r_1_2);
T (char, S (1), r_1_2);
T (char, S (2), r_1_2); // { dg-warning "\\\[-Wstringop-overflow"
"pr92814" }
+ size_t r_2_3 = UR (2, 3);
T (char, S (0), r_2_3);
T (char, S (2), r_2_3);
T (char, S (3), r_2_3); // { dg-warning "\\\[-Wstringop-overflow"
"pr92814" }
Martin, is there any harm in changing the testcase this way? The issue
reproduces with a cross and the two-line snippet at the top (just prune
the rest from the testcase). In case you want to have a closer look
why the warning is not issued on ppc64le (where I tested) vs. x86_64.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85316
[Bug 85316] [meta-bug] VRP range propagation missed cases
