fix(ci): make the toolchain version asserts diagnosable - #30
Merged
Conversation
`idris2 --version | grep -Fx 'Idris 2, version 0.7.0'` failed the job with no
output whatsoever: the pipeline swallowed the version, grep matched nothing,
and `set -o pipefail` ended the step before anything was printed. The log gave
no clue what version was actually installed.
Capture the version, echo it, then assert. The Idris2 check now matches on the
release ("0.7.0") rather than the whole string, since the digest-pinned
idris2-pack image reports a build suffix. The Zig check keeps its exact
equality but now reports what it found when it fails.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
hyperpolymath
added a commit
that referenced
this pull request
Jul 27, 2026
#31) Now that #30 made the assert diagnosable, the real mismatch is visible: ``` idris2 --version -> Idris 2, version 0.8.0-6ca00e72e ::error::expected Idris 2 0.7.0 ``` The digest-pinned `idris2-pack` image ships **0.8.0**, so the `0.7.0` assertion could never have matched — `CI / required` has never passed on its own terms. This is the last red check on `main`. ## Which side is wrong? Nothing in the repository actually requires 0.7.0: - `abi.ipkg` declares **no version constraint** — it depends only on `base`. - The `idris2 0.7.0` line in `.tool-versions` is **commented out**, so it pins nothing. - The only 0.7.0 references were the workflow asserts and their comments. So "0.7.0" was stale documentation that drifted from the digest. The **immutable digest is the authority**, and the version notes are brought into step with it rather than the reverse. The typecheck now actually runs against the toolchain that is genuinely installed. Also fixes the identical brittle assert in **`release.yml`**, which piped `idris2 --version` into `grep -Fx` under `pipefail` exactly as `ci.yml` did — it would have failed the same way, silently, on the next release. ## If 0.7.0 was intended Then the fix is the other direction: re-pin the container to a 0.7.0 image digest. Say so and I will swap it — but the current pair is contradictory either way. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 5 <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.
CI / requiredfailed on #28 atTypecheck the Idris2 ABIwith no output at all:The pipeline swallowed the version,
grep -Fxmatched nothing, andset -o pipefailended the step before anything reached the log — so the failure was undiagnosable
from CI alone.
Fix
Capture the version, echo it, then assert:
*"0.7.0"*) rather than the entire string.The digest-pinned
idris2-packimage reports a build suffix, so exact-stringequality could never have matched. The version pin is still enforced.
0.15.2) but now reports what it actually found.Both print their version on success too, so the log records which toolchain ran.
Follow-on from #28, which fixed the earlier failure in the same job (the image
ships no
xz, sosetup-zigcould not unpack its.tar.xz).🤖 Generated with Claude Code