Use case:

I have an app my-app and another app B. There's a bug that occurs only when B executes my-app. I want to directly debug my-app when executing B.

And BTW it's very painful to debug multi-process applications in gdb.

How:

• Add breakpoints on exec(e.g. break when exec my-app)
• Ask if launching external debugger on breakpoint hit
• Detach from the tracee to be debugged and launch external debugger(e.g. gdb in a new window) and let it attach to tracee.

Unresolved questions:

• Is this compatible with seccomp-bpf optimization?
• How could we transfer ptrace from tracexec to external debugger?
• Should the breakpoint be triggered on syscall enter stop or syscall exit stop?

Caveats:

• Further exec events will not be available for the detached tracee.

Alternative

Maybe I could implement gdb server protocol in tracexec and let gdb as a client connect to it. This way further exec events can also be traced.

• https://github.com/luser/rust-gdb-remote-protocol