The last one was gitea.com/gitea/runner, and this forge used two of its
packages: act/model to parse and plan workflows, act/exprparser to evaluate
expressions. Those plus the two small helpers they need are 3619 lines. The
other 12000 in the upstream act subtree run containers, which is the runner
process job, not the forge.
So github.com/hanzoai/act v1.0.0 is that subset, published alongside the
existing runner image in the same repo — the image and the library are both
"act", and the module path matches the repo.
Started at 13 gitea.com dependencies:
go-chi binding/cache/captcha/session in-tree
lunny/levelqueue replaced by ZapDB
lunny/dingtalk_webhook four structs, defined next to the sender
gitea/runner forked as hanzoai/act
...and the rest resolved with them.
Everything green; integration compile gate green.
Co-authored-by: Hanzo Dev <[email protected]>
zeekayandHanzo Dev
4f35e3fbbc