Baojun Wang

On AWS EC2 (t2.micro), it reports `cpuid_fault` is supported; however, it does nothing. hence even `arch_prctl(ARCH_SET_CPUID)` returns success, the system still cannot trap `cpuid` instruction. As a result, we won't...

Most of the samplePrograms compare the program's outputs, it might be beneficial to use a more general testing framework instead of using Make/bash. [tasty](https://github.com/feuerbach/tasty) sounds promising, It should support *MustSuccess*...

Had a brief talk with @RyanGlScott, we're fine with anything other than what we have right now :)

yes it's quite ugly, not to mention they're governed by the same `--real-proc` flag. OTOH I'm surprised docker just use host's `/etc/{passwd,shadow,group}`.

you're right, I must have saw something else irrelevant.

There are also other reasons, for instance, `/proc/stat` should depends on `get_nprocs` instead of using static values, we may not to expose all the interrupts and its counters.

Not any easier, I suppose we can add something like: `--mount src=mount_src,dst=mount_dst,type=xxx` just like docker.

Actually, I think using `-v,--volume` can be easier to parse with `cxxopts`, accepted flag(s): `-v x`, `-v x:y`, `--volume x`, `--volume x:y` `x` will be bind mounted to `y` inside...

docker filter certain syscalls such as `personality`, `seccomp`, `ptrace`..., see: https://docs.docker.com/engine/security/seccomp/#significant-syscalls-blocked-by-the-default-profile

There're more syscalls can set the flag, such as `pipe*`, `socket`, `memfd_create`, `timerfd_create`, `signalfd_create`, `fcntl`... pretty much anything returns a `fd`. FDs can live across `exec*`, however, most of the...