Viktor Malík

I agree here, cc @brendangregg who originally wrote the tool. On the other hand, as I already said, it's not that easy to check if a map is empty. Feel...

Also note that most "printing" actions such as `print(f)`, `time`, `join`, etc. are asynchronous operations, i.e. they are not performed inside the BPF program. Therefore, storing their results in maps...

> Hi, i wonder if you have considered a method like systemtap to generate c code, and then compile the c code to generate binary? Hi @chengshuyi, I'm not sure...

The main problem here is that distros don't use consistent kernel version numbering that can be relied on. It's true that we use something similar in [tests](https://github.com/iovisor/bpftrace/blob/master/tests/runtime/builtin#L164) but I'd be...

> Perhaps something like weak symbols for probes might make sense. In that a failure to attach is silent. Not sure if we'd want to support multiple weak attachments to...

The question is what would "tracing buffer" mean in this case. IIUC, buffers in dtrace are used for passing data from kernel to userspace, however, BPF has maps to do...

I'm not very familiar with DTrace, let alone with the speculative tracing feature but from what I understand, it's something that should be replaceable by using BPF maps. I'll try...

This doesn't seem like an issue with probe attachment rather than with process/program localization. @lenticularis39 could you perhaps have a look?

FWIW, this should not be related to DWARF, as simply attaching to a function in a binary should work even without DWARF. Could you post the output of `nm uprobe:/usr/lib/xorg/modules/input/libinput_drv.so`?...

Right, my bad. Should be `nm /usr/lib/xorg/modules/input/libinput_drv.so`.