Harness ends with DEADLYSIGNAL at start

[AdvenamTacet]

Often running a harness ends with:

AddressSanitizer: DEADLYSIGNAL
AddressSanitizer: DEADLYSIGNAL
AddressSanitizer: DEADLYSIGNAL
...
AddressSanitizer: DEADLYSIGNAL
AddressSanitizer: DEADLYSIGNAL
AddressSanitizer: DEADLYSIGNAL

In my experience, it always happens at the very beginning of running the harness.

While my setup is quite specific, I believe @mschwager observed the same behavior. (Am I right?)

[mschwager]

Yeah, I've seen this reliably a number of times. In my case, it would crash libFuzzer and the fuzzing process. This was from a while ago, so I don't immediately have the reproduction steps, but I did dig up these links that I was using to debug:

• https://stackoverflow.com/questions/77826203/addresssanitizer-randomly-throws-sigsegv-with-no-explanation
• https://stackoverflow.com/questions/76514542/inconsistent-output-from-addresssanitizer-intermittent-error-segv-on-unknown-ad
• https://groups.google.com/g/emscripten-discuss/c/QERnFWUT0_A
• Empty code randomly crashes with SEGV on unknown address google/sanitizers#1724
• Thread Sanitizer FATAL error on kernel version 6.6.6-x google/sanitizers#1716

I did try turning off ASLR, but that didn't stop the issue from happening for me.