Alastair Robertson

> In general, I'm a big fan of methods on objects BUT that's a big language change for bpftrace. If we added something like `@my_map.contains(mykey)` I think we'd also need...

Ok not really forever, I just meant it in the sense that while theoretically we can (and I'd say should!) deprecate and remove some things, in practice we've very rarely...

For anyone coming here from the changelog entry: `has_key` was implemented in PR #3485

Thanks for the report. For whatever reason, my machine is refusing to execute `__sock_sendmsg` so I haven't been able to reproduce the issue myself yet. ----- While typing this comment...

I can't reproduce this myself. I tried using bpftrace 0.20 and latest master and compiled my test program with gcc 11.5 and clang 20, using `-O0` (`-O3` caused the watchpoint...

I think objdump might be misleading here - it will only show you the instructions from the executable on-disk. bpftrace/BPF injects the breakpoint trap instruction when the executable is loaded...

Yeah, any kind of function can be called either with the traditional syntax or UFCS.

> I generally like this for the already stated purposes (readability, chaining, etc.) but we must realize that allowing this for all functions will allow things like: > > ```...

I'm fine without debuginfod - my only experience with it is it hanging bpftrace occasionally when it can't get a network connection! If people really do want it, we can...

Yeah it'll definitely help to make the problems in #3308 less common, but they'll still be there for people writing their own for-loops (and older kernels, as you say)