[breakertt]

This follows up on #119, as John (comment) and @Stuartwastaken (comment) suggested there. It adds these steps after the unit tests:

Step	What it does	Fails when
Build homa.ko (stripped)	make __STRIP__=1, with the same kernel as the normal build	the stripped module does not build
Unit tests, stripped (IPv6, IPv4)	the unit tests built with __STRIP__=1	a test fails
checkpatch	checkpatch on the change, in the module sources as they are (the files of make checkpatch)	checkpatch reports an error
checkpatch (stripped)	checkpatch on the change, in the files that go upstream after util/strip.py (the files of make checkpatch-net-next)	checkpatch reports an error

Both checkpatch steps call .github/scripts/checkpatch.sh, which runs checkpatch with the same options as the netdev patch checks (NIPA). The script checks only the lines that a pull request or push changes, because the existing code on main already has some checkpatch errors. A checkpatch error fails the step. Warnings and checks do not fail it. They appear as warnings in the summary of the CI run.

The script is in .github/scripts/, where other projects keep the scripts of their workflows, for example libbpf, PyTorch and Neovim. You can also run it locally, before you push: .github/scripts/checkpatch.sh <base> <linux> [stripped].

I tested it in my fork with three pushes:

Push	Run	Stripped build and unit tests	checkpatch	checkpatch (stripped)
this commit, on a new branch	40660538429	pass	skipped (new branch, nothing to compare with)	skipped (new branch, nothing to compare with)
one CHECK in homa_qdisc.c	40660800474	pass	pass, one annotation	skipped (the file does not go upstream)
four ERRORs in homa_pool.c	40661091120	build fails (the test function has no prototype)	fails	fails

[Stuartwastaken]

Looks good, Tianyi! One thing for later, not this PR: __STRIP__=1 strips with the preprocessor, so it can’t catch mistakes in what util/strip.py itself writes. Building that output against the same 6.17.8 tree found #134, and with -Werror it also catches the #126 dangling else. I have that as a CI step and can rebase it onto this once it’s merged, if that works for you.

[breakertt]

Hi Stuart, That sounds good to me, thank you! Cheers, Tianyi

…

________________________________ From: Stuart Ray ***@***.***> Sent: Wednesday, 28 October 2026 19:35:15 To: PlatformLab/HomaModule ***@***.***> Cc: breakertt ***@***.***>; Author ***@***.***> Subject: Re: [PlatformLab/HomaModule] Add a stripped build and checkpatch to CI (PR #133) [https://avatars.githubusercontent.com/u/88628337?s=20&v=4]Stuartwastaken left a comment (PlatformLab/HomaModule#133)<#133 (comment)> Looks good, Tianyi! One thing for later, not this PR: STRIP=1 strips with the preprocessor, so it can’t catch mistakes in what util/strip.py itself writes. Building that output against the same 6.17.8 tree found PlatformLab/HomaModule#134 · Strip the unit-test side of…<#134>, and with -Werror it also catches the #126<#126> dangling else. I have that as a CI step and can rebase it onto this once it’s merged, if that works for you. — Reply to this email directly, view it on GitHub<#133?email_source=notifications&email_token=AEHMMGHQAEHTETBBAALVETT5WJC7HA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTMNBXHA3DSNBSG43KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-6478694276>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AEHMMGFN7ENDCGURZTDTMQD5WJC7HAVCNFSNUABFKJSXA33TNF2G64TZHMYTGOJWGM3DANZXHNEXG43VMU5TMMBXGYYDIOBZHE32C5QC>. You are receiving this because you authored the thread.Message ID: ***@***.***>