Alastair Robertson

I've done some work on this one already (in the headers branch). I think all the BPF/LLVM work is done, it just needs to use libclang to parse the header...

My current implementation will end up looking like your second example there - the first one would be a bit harder as bpftrace would have to know what the arguments...

Yeah this looks good to me!

There's not currently any nice way of getting those values, but it's definitely something that can be worked on

@lxyscls I'm not sure if these issues are related, but for the https://github.com/ajor/bpftrace/issues/50 one I've proposed a fix here: https://github.com/iovisor/bpftrace/issues/2292

Definitely a good feature, but I'm not too keen on the proposed syntax. What about something like this as an alternative: ``` delete(@work[_, 10]); // deletes @work[10, 10] and @work[1,...

I actually liked the factoring out of `prepareFormatStringDataMap()` you had in your previous version! It was nice to not have that section of duplicated code. Any reason you removed it?

Is there an example use-case for this?

Related to #68 (func for kretprobes) and #101 (stacks for retprobes).

The `get_func_ip` helper which we used to fix `func` in kretprobes doesn't work for uretprobes 😢 Some discussion here: https://lore.kernel.org/bpf/ZKuyKj8jrEPzWYMM@krava/T/#u There seems to be an agreement to change the behaviour...