Robert O'Callahan

> Where can I read more about "monkeypatch"? Monkeypatcher.cc > Do you mean instead of the assert a plain check should be done and then a different code be executed...

> Yes. Should we query the kernel version and make that conditional? No, I don't want to add stuff for a kernel version we don't officially support. I mean, maybe...

I'll add info to #2929.

> `PTRACE_SETSIGMASK` is available with kernel 3.11 - are you sure that the current task tearddown code would work without the suggested changes with a 3.11 kernel? If not, then...

> E.g. https://buildkite.com/julialang/rr/builds/220#7dd3d2e4-d59f-4d14-a731-24ccf24ff78e Was this the correct link? There I see ``` 2747/2761 Test #2448: rseq-32 ........................................................... Passed 58.58 sec ```

``` [FATAL /cache/build/exclusive-amdci1-0/julialang/rr/src/PerfCounters.cc:630:read_ticks()] (task 47041 (rec:65972) at time 355) -> Assertion `!counting_period || interrupt_val

i.e. this `SCHED` is not in the rseq critical section so I don't know why this would trigger anything abnormal.

I think f2faffc50d351b14718e47abffd4ac6f75b61ea3 should have fixed this.

Er no, not this assertion failure.

``` [FATAL /workspace/srcdir/rr/src/Task.cc:2822:read_bytes_helper() errno: EIO] (task 20367 (rec:20367) at time 1348590) -> Assertion `false' failed to hold. Should have read 48 bytes from 0x7f6b8c7534e8, but only read -1 [FATAL /workspace/srcdir/rr/src/log.cc:414:emergency_debug()]...