https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92026
Martin Sebor <msebor at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed| |2019-10-08
Component|testsuite |middle-end
Assignee|unassigned at gcc dot gnu.org |msebor at gcc dot
gnu.org
Ever confirmed|0 |1
--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> ---
The code from the test case on line 159 in Wstringop-overflow-18.c:
char a8[8], b4[4];
void f (void)
{
__builtin_strcpy (b4, "123");
char *d = a8 + 5;
__builtin_memcpy (d, b4, 4);
}
is represented like this on the failing sparc-sun-solaris2.11 (same on
mips64el-linux):
<bb 2> [local count: 1073741824]:
MEM <unsigned char[4]> [(char * {ref-all})&b4] = MEM <unsigned char[4]>
[(char * {ref-all})"123"];
MEM <unsigned char[4]> [(char * {ref-all})&a8 + 5B] = MEM <unsigned char[4]>
[(char * {ref-all})&b4];
return;
but like this on the passing arm-none-eabi:
<bb 2> [local count: 1073741824]:
MEM <unsigned char[4]> [(char * {ref-all})&b4] = "123";
MEM <unsigned char[4]> [(char * {ref-all})&a8 + 5B] = "123";
return;
and like this on the passing x86_64:
<bb 2> [local count: 1073741824]:
MEM <unsigned char[4]> [(char * {ref-all})&b4] = MEM <unsigned char[4]>
[(char * {ref-all})"123"];
MEM <unsigned int> [(char * {ref-all})&a8 + 5B] = 3355185;
return;
The warning knows how to handle all three forms but it doesn't know handle the
first case correctly: it doesn't set the size of the store in the failing case.
