Delyan Kratunov

@xh4n3 yeah, not surprised, this is similar to an internal use case we had that hit this issue. There we split the printfs into multiple calls but that's not a...

> The idea is to be a minimal compiler backend that can still produce reasonable BPF bytecode. I think this issue is vastly understating how difficult such a backend is....

> As a matter of that, we could also provide a script which would fill the necessary docker run parameters, especially mount the necessary volumes. I would rather not, that's...

Someone internally pointed me to [this blog post](https://blog.galowicz.de/2019/04/17/tutorial_nix_cpp_setup/) which, apart from the FP constructions that twist your brain, does exactly the type of matrix build that we want (we'd have...

I spent some time messing around with Nix. Once you wrap your brain around its world view, our model becomes easy to express. E.g., this uses the upstream nixpkgs (should...

#2396 also has the story of needing to pin bpftool versions - something that nix would make easy in any environment. Alternatively, we can vendor bpftool I guess.

Here's a fun one! `bpftrace_test` tests can be **dramatically** faster: Without root: ``` $ time gtest-parallel tests/bpftrace_test [1/525] utils.get_cgroup_hierarchy_roots (55 ms) Note: Google Test filter = utils.get_cgroup_hierarchy_roots [==========] Running 1...

I'm fighting with it in https://github.com/iovisor/bpftrace/pull/2370/commits/4eb280b07ee683f62176e06bdbb594c146a72379 but I can't make it stable yet. Now the pidof invocation to get the child pid after it has started is failing. I may...

Unfortunately, I don't think it was fully fixed, just made a bit better. I still don't fully understand how this happens (and why it happens on _that_ test specifically) but...

Okay, I give up. #2414 doesn't help either, so this is not about the test runner racing the BEFORE calls. Maybe it's something to do with --usdt-file-activation and whatever kernel...