A branch with no PR can still be checked on demand - #23
Merged
Merged
Conversation
Narrowing `push` to `main` stopped the double run for every commit on a PR branch, and took away the only way a branch with no PR yet ever got checked. `workflow_dispatch` hands that back — haus's check.yml has had it all along, and it is the trigger block workshop's docs/ci.md rule 2 prescribes. The factory opens a PR for everything here, so this is rarely the path that gets used; it is the escape hatch for a branch pushed without one. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019v1reBfZd8d4euMNyXry6B
JulienMartel
added a commit
to hausfold/workshop
that referenced
this pull request
Sep 12, 2026
…e wrote down Two leftovers from the CI pass. Narrowing `push` to `main` took away the only way a branch with no PR yet ever got checked. haus's check.yml already hands that back with `workflow_dispatch`; workshop's test.yml did not, and neither did the canonical trigger block in docs/ci.md's rule 2. Both now match haus. factory's test.yml had the same gap and takes the same one line in hausfold/factory#23. The path-filter question is now measured and closed. Against a real ignore list over each repo's last 40 merged PRs, the share that would skip is haus 1, trill 6, scruff 10, perch 11, nebelung 12 — two or three minutes saved on a quarter of PRs at best. Not enough to buy a second silent copy of "which files does this job protect?", which rots; and scruff's release path reaches its gate through `workflow_call`, so a filter that skips a job on a docs path skips it for a tag too. It goes under "What we deliberately don't do" with the numbers and with the method, because the loose version of that measurement (which says 80%) is how the idea keeps coming back. The same section's "no hosted binary cache" now says what it does not rule out: a public read-only substituter for something nixpkgs doesn't carry needs no account and no token, and fails soft to building from source. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019v1reBfZd8d4euMNyXry6B
JulienMartel
added a commit
to hausfold/workshop
that referenced
this pull request
Sep 12, 2026
…e wrote down (#581) * A branch with no PR can still be checked, and path filters are a no we wrote down Two leftovers from the CI pass. Narrowing `push` to `main` took away the only way a branch with no PR yet ever got checked. haus's check.yml already hands that back with `workflow_dispatch`; workshop's test.yml did not, and neither did the canonical trigger block in docs/ci.md's rule 2. Both now match haus. factory's test.yml had the same gap and takes the same one line in hausfold/factory#23. The path-filter question is now measured and closed. Against a real ignore list over each repo's last 40 merged PRs, the share that would skip is haus 1, trill 6, scruff 10, perch 11, nebelung 12 — two or three minutes saved on a quarter of PRs at best. Not enough to buy a second silent copy of "which files does this job protect?", which rots; and scruff's release path reaches its gate through `workflow_call`, so a filter that skips a job on a docs path skips it for a tag too. It goes under "What we deliberately don't do" with the numbers and with the method, because the loose version of that measurement (which says 80%) is how the idea keeps coming back. The same section's "no hosted binary cache" now says what it does not rule out: a public read-only substituter for something nixpkgs doesn't carry needs no account and no token, and fails soft to building from source. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019v1reBfZd8d4euMNyXry6B * The path-filter no is about the denylist, and it says which list it measured Assurance pass on the first commit found three things worth the rewrite. The scruff argument was wrong for the mechanism the heading named. A `paths-ignore:` under `on:` cannot skip anything at a tag — a `workflow_call` invocation does not evaluate the called workflow's own triggers, so release.yml still runs the full gate. The risk is real only for an in-job changed-files filter, which IS evaluated on that run, so the stanza now names that implementation and says where a filter would have to go if one were ever worth it. The section also contradicted the hausfold.co paragraph above it, which calls a `paths:` filter the first thing to reach for. Allowlist and denylist were never distinguished: a `paths:` says what one task workflow is about, a `paths-ignore:` on the gate says what the repo merges unchecked. Only the second is the no, and the heading now says so. The numbers were unreproducible because the ignore list they were measured against was nowhere in the doc. It is in the doc now, and the closing paragraph points at it instead of at an "actual list you would ship" the reader would have to invent. Two smaller ones: the substituter carve-out was answering "a cache we run" when the paragraph above rules out one we rent and push to, and fail-soft is the property that makes the difference, so it says that; and rule 2's block now says it is the floor rather than the whole `on:`, since scruff's gate also needs `workflow_call`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019v1reBfZd8d4euMNyXry6B --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
One line on
.github/workflows/test.yml's trigger block:workflow_dispatch:.Why
The CI pass narrowed
pushtobranches: [main]so a commit on a PR branchstops being checked twice, once as
pushand once aspull_request. That isright, and it took something with it: a branch pushed without a PR now gets
no run at all.
workflow_dispatchis the replacement. haus'scheck.ymlhas had it allalong, and workshop's
docs/ci.mdrule 2 prescribes it as part of thecanonical trigger block (that rule's own yaml was missing it too — fixed in the
workshop PR this one pairs with).
Verify
Then: Actions → Tests → Run workflow offers a branch picker, and a PR here
still produces exactly one run per commit.
Watch out
factory shiftreadsgh run list -b main, so it is unaffected either way —nothing here reads a run on a non-main branch. The factory opens a PR for
everything in this repo, so the dispatch button is an escape hatch, not a path
anyone walks daily.
🤖 Generated with Claude Code
https://claude.ai/code/session_019v1reBfZd8d4euMNyXry6B