Determinism tests fail if ASLR can't be disabled #3460
Description
The determinism tests are failing for me. They have been failing for a while:
$ ./setup test determinism
[...]
The following tests FAILED:
42 - determinism1-compare-shadow (Failed)
46 - determinism2-shadow-compare (Failed)
determinism1:
$ diff /shadow/build/src/test/determinism/determinism1a-shadow.data/hosts/testnode1/test-determinism.1000.strace /shadow/build/src/test/determinism/determinism1b-shadow.data/hosts/testnode1/test-determinism.1000.strace
12c12
< 00:00:01.000000000 [tid 1000] ioctl(3, 21505, 140732830198864) = -25 (ENOTTY)
---
> 00:00:01.000000000 [tid 1000] ioctl(3, 21505, 140736331501840) = -25 (ENOTTY)
17c17
< 00:00:01.000000000 [tid 1000] ioctl(3, 21505, 140732830198864) = -25 (ENOTTY)
---
> 00:00:01.000000000 [tid 1000] ioctl(3, 21505, 140736331501840) = -25 (ENOTTY)
determinism2:
$ diff /shadow/build/src/test/determinism/determinism2a-shadow.data/hosts/peer1/test-phold.1000.strace /shadow/build/src/test/determinism/determinism2b-shadow.data/hosts/peer1/test-phold.1000.strace
26c26
< 00:00:01.000000000 [tid 1000] futex(<pointer>, 129, 2147483647, <pointer>, <pointer>, 401523232) = 0
---
> 00:00:01.000000000 [tid 1000] futex(<pointer>, 129, 2147483647, <pointer>, <pointer>, 408764464) = 0
These look like pointers (but we don't know they're pointers because of the ioctl and futex syscalls).
Earlier I see:
00:00:00.194137 [70289:shadow] n/a [WARN] [n/a] [shadow.rs:180] [shadow_rs::shadow] Could not disable address space layout randomization. This may affect determinism: ENOSYS: Function not implemented
Stack backtrace:
0: anyhow::error::<impl core::convert::From<E> for anyhow::Error>::from
at /home/user/.cargo/registry/src/index.crates.io-6f17d22bba15001f/anyhow-1.0.89/src/backtrace.rs:27:14
1: shadow_rs::shadow::run_shadow
at /shadow/src/main/shadow.rs:178:11
2: main_runShadow
at /shadow/src/main/shadow.rs:473:22
3: main
at /shadow/src/main/main.c:17:12
4: __libc_start_call_main
5: __libc_start_main@@GLIBC_2.34
6: _start
It seems likely that ASLR isn't being disabled, so the pointers aren't deterministic in the strace.
Maybe we should make our determinism test better handle the case where ASLR can't be disabled.