Robert O'Callahan

> @Martiusweb speculates that rr only needs ptrace, perf_event_open and maybe also userfaultfd, and he suggests that we try enabling these first, and iterate on any missing syscalls afterwards. That...

I don't think you need to worry about `clone`. rr does not create new namespaces (though it supports recording children that do). It uses `clone` to create processes and threads,...

Though you might need to whitelist them.

> That's interesting, is it related with this commit? torvalds/linux@58d0a86 As the commit message says, in the past a process could use ptrace() to negate any seccomp() policy imposed on...

The addition of `DW_FORM_GNU_ref_alt` and `DW_FORM_GNU_strp_alt` mean that gimli can't parse these debug binaries at all.

It looks like this hasn't actually been standardized in DWARF5 yet. I thought perhaps DWARF5 DWO would cover these use-cases but as far as I can tell it doesn't.

Aha, so it is.

I guess the main thing we need to do for gimli is to extend `AttributeValue` to support alt-references for `DebugStrRef` and `DebugInfoRef`. We could add new `AttributeValue` cases, e.g. `DebugStrRefAlt`/`DebugInfoRefAlt`,...

I filed #288 to add `DebugStrRefSup` and `DebugInfoRefSup` to `AttributeValue`.

If you do `rr replay -a` without writing the results to a log file you'll get instructions for starting the emergency debugger. Try following those instructions and then doing "where"...