Hi,
I filed this on bugzilla: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59744 but I thought I'd mention it here too.
This slightly strangely written program (it's distilled down from frame_offset_overflow in the gcc source itself) should print "bigger" if the first argument is bigger than 10 (or negative, but let's ignore that please):
#include <stdlib.h> #include <stdio.h>
int a[2] = { 10, 20 };
int is_bigger (long offset, int index) { unsigned long size = -offset;
if (size > a[index]) { printf("bigger\n"); return 1; }
return 0; }
int main (int argc, char** argv) { long v; v = atol(argv[1]); is_bigger(-v, 0); return 0; }
When compiled at -O1 or above (and with inlining disabled at -O2 and above), though, it bungles the 0 case:
(t-doko)mwhudson@arm64:~$ gcc-4.9 -O3 test.c -o test -fno-inline -Wall (t-doko)mwhudson@arm64:~$ ./test 1 (t-doko)mwhudson@arm64:~$ ./test 11 bigger (t-doko)mwhudson@arm64:~$ ./test 0 bigger (t-doko)mwhudson@arm64:~$ gcc-4.9 -O0 test.c -o test -Wall (t-doko)mwhudson@arm64:~$ ./test 1 (t-doko)mwhudson@arm64:~$ ./test 11 bigger (t-doko)mwhudson@arm64:~$ ./test 0 (t-doko)mwhudson@arm64:~$
What's going on? Here's the disassembly of is_bigger (at O3):
0000000000400608 <is_bigger>: 400608: b0000082 adrp x2, 411000 <_GLOBAL_OFFSET_TABLE_+0x28> 40060c: 91010042 add x2, x2, #0x40 400610: a9bf7bfd stp x29, x30, [sp,#-16]! 400614: 52800003 mov w3, #0x0 // #0 400618: 910003fd mov x29, sp 40061c: b8a1d841 ldrsw x1, [x2,w1,sxtw #2] 400620: ab00003f cmn x1, x0 400624: 540000a2 b.cs 400638 <is_bigger+0x30> 400628: 90000000 adrp x0, 400000 <_init-0x3f8> 40062c: 911b6000 add x0, x0, #0x6d8 400630: 97ffff90 bl 400470 puts@plt 400634: 52800023 mov w3, #0x1 // #1 400638: 2a0303e0 mov w0, w3 40063c: a8c17bfd ldp x29, x30, [sp],#16 400640: d65f03c0 ret
Basically it seems that the condition "-offset > val" is being compiled as "val + offset does not overflow", which is not valid for offset == 0.
This seems to me to be the underlying cause behind https://bugs.launchpad.net/ubuntu/+source/gcc-4.8/+bug/1263806 ("gccgo fails to compile tomb.go on arm64").
Cheers, mwh