Delyan Kratunov

Does the fact that the SDK now ships with a maven repository for the support library (which, frankly, I find to be an awful hack) change the project's stance on...

Yeah, the store is wrong of course. This assumption needs unwinding, that whole ptrtoint cast would now be wrong: https://github.com/iovisor/bpftrace/blob/833ca149e8e1f41a0d8b13826f8261c1b75673a4/src/ast/passes/codegen_llvm.cpp#L2090-L2097 > In fact, the only problem I can see here...

Okay, so that counter-example is useful to expose the fundamental confusion that I've been struggling to put my finger on. When instantiating the clang frontend and parsing types, we don't...

We should probably fix that branch and re-release 0.16.1 but fwiw a higher libbpf version will soon be required. #2370 changes this code to assume at least libbpf 0.8.1.

> Which one is it? It's a bit ambiguously worded. I meant it more as "use bpf_object to discover the maps and programs but do relocation and loading the way...

#2321 got a little stuck on dealing with old libbpf versions, so I decided to hurry up and do the import into the repo. Uhm, that hasn't gone according to...

Updates in this reroll: 1. User-facing documentation 2. Codegen tests updated for both debug info and delete_filter 3. Semantic analyzer and runtime tests 4. clang-format every commit 5. Addressed most...

@viktormalik what's the best way to reproduce the CI environment? The Vagrantfile seems out of date (a bunch of the images are not in vagrantcloud anymore). Is there a convenient...

Okay, I've hit a bit of a roadblock with the BTF support in this PR. Namely, LLVM only learns about BTF in version 10. I can definitely gate the runtime...

> we run CI for 9+ only, so we basically don't support anything below. Yes, I also want to clean that up in the code, I don't think we can...