Actually, this problem can be reproduced with an even simpler program:
#include <stdio.h>
#include <unistd.h>
#include <sys/time.h>
int main() {
struct timeval tv;
gettimeofday(&tv, NULL);
}
If I compile this with `gcc -O2 -o test_O2 test.c` and run
`libunwind/tests/test-ptrace -s -v ./test_O2` to trace it, I will get the
same behavior as above with `main+0x1e` showing up twice in the backtrace.
However, the weird thing is if I replace gettimeofday() call with a
sleep(1) call, even compiling with the `-O2` flag, backtrace looks fine.
Maybe there's a strange interaction between libunwind and gettimeofday
syscall?
-
Hai
On Tue, Aug 15, 2017 at 9:52 PM, hai huang <[email protected]> wrote:
> I have a simple test program (test.c) to reproduce this problem:
>
> #include <stdio.h>
> #include <sys/time.h>
>
> void g() {
> struct timeval tv;
> gettimeofday(&tv, NULL);
> }
>
> void f() { g(); }
>
> int main() {
> f();
> }
>
> If I compile this program with `gcc -O0 -o test_O0 test.c`, and do
> backtrace by running `libunwind/tests/test-ptrace -s -v ./test_O0`, I get
> the expected results when gettimeofday() is intercepted, see below:
>
> 00007ffed67fee89 +0x432 (sp=00007ffed67fb000)
> 00000000004005be g+0x28 (sp=00007ffed67fb010)
> 00000000004005e3 f+0xe (sp=00007ffed67fb070)
> 00000000004005f4 main+0xe (sp=00007ffed67fb080)
> 00007fbda5924830 __libc_start_main+0xf0 (sp=00007ffed67fb090)
> 00000000004004c9 _start+0x29 (sp=00007ffed67fb150)
> ================
>
> I slightly modified the output format of test-ptrace.c so the output is
> more succinct. But this shows that when gettimeofday is invoked, the
> backtrace is main() -> f() -> g() -> somewhere in libc -> gettimeofday(),
> which is expected.
>
> However, when I compile the same program with `gcc -O2 -o test_O2 test.c`,
> and do backtrace by running `libunwind/tests/test-ptrace -s -v ./test_O2`,
> I get the following:
>
> 00007ffda5dc4e89 +0x432 (sp=00007ffda5ca3560)
> 00000000004004be main+0x1e (sp=00007ffda5ca3570)
> 00000000004004be main+0x1e (sp=00007ffda5ca35a0)
> 00007fec46b15830 __libc_start_main+0xf0 (sp=00007ffda5ca35d0)
> 0000000000400509 _start+0x29 (sp=00007ffda5ca3690)
> ================
>
> Well, since f() and g() don't do anything other than eventually calling
> gettimeofday(), I can understand the compiler got rid of these unnecessary
> function calls and just put the invocation of gettimeofday() inside of
> main. I confirmed this by dissembling the compiled binary. However, what I
> don't understand is why `main+0x1e` appears twice in the backtrace.
>
> I'm on a Linux x86_64 Ubuntu 16.04 machine. Any help is appreciated!
>
> Thanks!
>
> -
> Hai
>
_______________________________________________
Libunwind-devel mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/libunwind-devel
