Robert O'Callahan

That looks like you're running rr under rr somehow...

I rebuilt your Gecko tree at revision f50fadf05f3959a021395ac67fb821083cf5bdc9 and tried following the steps of comment https://github.com/mozilla/rr/issues/1771#issuecomment-240763610. Tabs crashed a lot and the replay worked fine. This is with gcc 6.1.1....

My version of clang is 3.8.0. I'll try that...

Basically the same results with clang 3.8.0 here. Note that these are opt builds of rr and Firefox. (`cmake -DCMAKE_BUILD_TYPE=Release` for rr.)

Did you try running the rr tests with your clang build of rr?

@glandium can maybe help you with zen_workaround.

You need to run it as root.

You may wish to retry with master. Just yesterday I landed some improvements in this area.

https://github.com/rr-debugger/rr/commit/e54a6f9324c683498aabaee7abd60d3cf6fe73f0 prints the full registers if we catch this assertion again.

This is the rr stack: ``` #6 0x0000558bd76e2bf3 in rr::RecordSession::process_syscall_entry (this=0x558bd88208e0, t=0x558bd8831f00, step_state=0x7ffce960271c, step_result=0x7ffce96027e0, syscall_arch=rr::x86_64) at /cache/build/exclusive-amdci3-0/julialang/rr/src/RecordSession.cc:1854 #7 0x0000558bd76dccc3 in rr::RecordSession::handle_seccomp_traced_syscall (this=0x558bd88208e0, t=0x558bd8831f00, step_state=0x7ffce960271c, result=0x7ffce96027e0, did_enter_syscall=0x7ffce960271b) at /cache/build/exclusive-amdci3-0/julialang/rr/src/RecordSession.cc:430 #8 0x0000558bd76dd99a...