Remove Rust/Aya-based toolchain #869
Conversation
|
Could you run some end-to-end testing with Parca Demo (or some other test program you trust) to double-check that everything works fine? 😄 |
kakkoyun
force-pushed
the
revert_rust_toolchain
branch
from
686f313 to
de53aba
Compare
I have already done that. We do another run. After see everything ok with actions. |
|
I'll check the failures soon. |
kakkoyun
force-pushed
the
revert_rust_toolchain
branch
3 times, most recently
from
1d28fee to
9f91aeb
Compare
|
The CodeQL action is failing, but seems we can remove the LLVM bitcode generation (see below), and just generate the BPF object code |
|
|
||
| .PHONY: clean | ||
| clean: | ||
| cargo clean | ||
| rm -f $(OUT_BPF) |
There was a problem hiding this comment.
nit: What do you think about also adding - rm -rf target/ so when people rebase on this branch the targets/ directory that might have been created by the previous build is removed, too?
| -xc \ | ||
| -nostdinc \ | ||
| -target bpf \ | ||
| -O2 -emit-llvm -c -g $< -o $(@:.o=.ll) |
There was a problem hiding this comment.
We can do this in one step without requiring the external linker binary
something like:
| -O2 -emit-llvm -c -g $< -o $(@:.o=.ll) | |
| -O2 -c -g $< -o $(@:.o) |
There was a problem hiding this comment.
Have we already used any branches in the agent? or any other projects?
| -nostdinc \ | ||
| -target bpf \ | ||
| -O2 -emit-llvm -c -g $< -o $(@:.o=.ll) | ||
| $(CMD_LLC) -march=bpf -filetype=obj -o $@ $(@:.o=.ll) |
There was a problem hiding this comment.
And we wouldn't need llc
Signed-off-by: Kemal Akkoyun <kakkoyun@gmail.com>
Signed-off-by: Kemal Akkoyun <kakkoyun@gmail.com>
kakkoyun
force-pushed
the
revert_rust_toolchain
branch
from
9f91aeb to
72b552d
Compare
Signed-off-by: Kemal Akkoyun kakkoyun@gmail.com
This PR reverts the changes that introduced in #377.
We are basically removing the Rust, Aya-based toolchain for the time being to be able to deliver our goals quicker. Aya has some rough edges such as aya-rs/aya#351, we were missing certain APIs to be able to write complex eBPF programs (programs to solves issues such as #768).
This being said, we will still continue to spent our innovation tokens on Rust and Aya for the further. The team still wants to pursue the path when the conditions are ripe.