TSAN issue: ThreadSanitizer:DEADLYSIGNAL

[nicola-cab]

[2023/12/06 11:01:12.367] 1: ThreadSanitizer:DEADLYSIGNAL
[2023/12/06 11:01:12.367] 1: ThreadSanitizer:DEADLYSIGNAL
[2023/12/06 11:01:12.367] 1: ThreadSanitizer: nested bug in the same thread, aborting.
[2023/12/06 11:01:12.482] 1/1 Test #1: CoreTests ........................***Failed   96.64 sec
[2023/12/06 11:01:12.482] 0% tests passed, 1 tests failed out of 1

https://spruce.mongodb.com/task/realm_core_stable_macos_1100_arm64_tsan_core_tests_pat[…]6570528261837d6f32012af9_23_12_06_10_52_51/logs?execution=0

Core version

Core version: 13.24.0

[kiburtse]

It is not specific to any test. It started to happen right after #6911. We essentially started to run with sanitizers on production like fully optimized builds, and this builder macos_1100_arm64 is what was added. It wasn't tested before with sanitizers at all. This error happens on many tests in pure core and sync test suites only. After searching a bit on this problem, it looks to as compiler and sanitizers implementation issue. I don't think that we can do anything for this other than update the testing platform and the compiler.

[nicola-cab]

OK, the only problem I have is that these checks (that are needed) if they can fail spuriously, will make releases unpredictable sometimes, or in general trigger false positives. Ideally, we want to avoid re-running things for false positives (does the same apply to evergreen and Jenkins).
However, if you think there is no solution. Probably I can close this.

[nicola-cab]

This PR #7206 has disabled the builder.

[kiburtse]

since #7383 reenabled these tests, it's the problem once again

I've put the fix to run this builder on newer macos13 here #7472 as a possible solution

[nicola-cab]

thanks @kiburtse

[sync-by-unito]

➤ PM Bot commented:

Jira ticket: RCORE-1906