Add CI job that runs unit tests under rr's chaos mode #18979

issue practicalswift opened this issue on May 15, 2020

practicalswift commented at 5:52 AM on May 15, 2020: contributor

As noted in #16337 (comment) rr's chaos mode can be used to trigger some classes of intermittent bugs in unit tests (such as bugs related to scheduling non-determinism).

It would be nice to have a CI job running under chaos mode to increase the likelihood of us finding such issues.

I think this is a "good first issue" :)

Context for those new to rr and its "chaos mode":

rr aspires to be your primary C/C++ debugging tool for Linux, replacing — well, enhancing — gdb. You record a failure once, then debug the recording, deterministically, as many times as you want. The same execution is replayed every time. -- https://rr-project.org/

Chaos mode to make intermittent bugs more reproducible. -- https://robert.ocallahan.org/2016/02/introducing-rr-chaos-mode.html

Personally I think rr is a fantastic piece of software which is right up there with projects such as libFuzzer and the sanitizers (ASan, MSan, TSan, UBSan) when it comes to its potential to help us produce robust and secure software. Highly recommended!
chrdevmar commented at 8:20 AM on May 15, 2020: none

keen to give this a crack
fanquake added the label Tests on May 15, 2020
MarcoFalke commented at 12:16 AM on May 16, 2020: member
I don't think this works. Is there any ci provider out there that comes with rr support?
```
dmesg | grep PMU
[    0.519557] Performance Events: unsupported p6 CPU model 63 no PMU driver, software events only.
```
MarcoFalke commented at 5:53 PM on May 19, 2020: member

Closing for now. rr is only supported on specific intel CPUs. AMD ryzen or any virtual CPU is not supported.
MarcoFalke closed this on May 19, 2020
DrahtBot locked this on Feb 15, 2022

Contributors

Labels

Linked (view graph)

#16337 Intermittent (and rare) unit test failure in blockfilter_index_tests/blockfilter_index_initial_sync