Joseph Devietti

@rrnewton I think this isn't done yet, we are still letting a lot of nondeterminism seep in via `/proc`. I'd feel more comfortable with having a properly-empty `/proc` and adding...

Signal 6 is SIGABRT: we support a process sending SIGABRT to itself (as part of the `abort` function). So this may be some other SIGABRT idiom. Perhaps we can support...

I like this approach a lot. I think ultimately for this next system we will need to handle threads, and sequentialize them. However, we could start by not supporting threads...

Starting with a fixed amount of free space seems like a great solution, `df` will always naturally return a reproducible value then.

Just to cast blame appropriately here, I think this behavior is the result of my _extremely_ half-baked implementation of `mkstemp` et al. These system calls should choose names that do...

I think we should take a greedy approach - do the lowest-hanging fruit first, which is deadlocks and `timer_create`. Then we'll re-run dettrace builds and see how the errors shift....

We would crash if we saw a `perf_event_open` syscall, so `rdpmc` is not of concern :-) Closing this.

If I understand correctly, when the tracee calls `unlink`, we know that we want to inject a `newfstatat` and update some monitor state with the `newfstatat` result, and then go...

I think this issue is worth solving in a clean way, since it will come up a lot when building software, and the solution is non-trivial. The "modified in the...

Reopening this until we are actually catching EINTR and throwing a `runtime_error`. Probably should do this in a blanket way for all system calls.