ci: DRIFT is advisory here — python is the reference, not a port - #91
Merged
Conversation
For the nine ports DRIFT is a hard gate, and rightly so: it asks whether the port still matches the reference, and a port can be wrong about that. This repo IS the reference. python_signatures.json is DERIVED from this tree, so python cannot drift from itself. A difference after regeneration means only that the snapshot in porting-sdk has not caught up yet — which is bookkeeping downstream of this repo, not a defect in it. Failing python's CI on that inverted the dependency. Adding public API here — an ordinary thing to do in the reference implementation — required a lockstep change in another repo before python could go green, so the source of truth was blocked on its own derivative. That is why a release sat unpublishable while a coordinated pass was in flight. The gate now reports the delta and returns 0. The delta is worth printing: it is exactly the new surface, and exactly what gets regenerated downstream after this merges. Nothing is weakened for the ports. They still check against the reference with a hard DRIFT gate. This tree is still enumerated on every run by SIGNATURES, which still fails loudly if it cannot be enumerated, and which SEMVER-DIFF still depends on — so a version bump that does not match the surface change is caught here, as before. Verified both paths: clean snapshot reports "matches this tree"; with new public API present it prints the shortstat and passes. Full run-ci otherwise unchanged.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Makes DRIFT advisory in this repo only, so python can merge and publish without a lockstep change in porting-sdk.
The inversion
For the nine ports, DRIFT is a hard gate and should be: it asks whether the port
still matches the reference, and a port can be wrong about that.
This repo is the reference.
python_signatures.jsonis derived from thistree. Python cannot drift from itself — a difference after regeneration means
only that the snapshot in porting-sdk has not caught up.
Gating python's CI on that inverted the dependency. Adding public API here — an
ordinary thing to do in a reference implementation — turned python red until
someone landed a change in another repo. The source of truth was blocked on its
own derivative. That is why
v3.3.0has been sitting unpublishable while acoordinated pass was in flight, and why every recent surface change here needed
a paired porting-sdk PR before it could go green.
What changes
DRIFT prints the delta and returns 0:
The delta is printed because it is genuinely useful — it is exactly the new
surface, and exactly what gets regenerated downstream after this merges.
What does not change
cannot be enumerated at all.
surface change is still caught here.
The only thing removed is python failing because another repo's file is stale.
Verified
Both paths exercised: with a clean snapshot it reports "matches this tree"; with
new public API present (replayed a real +682-line regen) it prints the shortstat
and passes. Full
run-ciotherwise unchanged — the only reds are TYPECHECK andTEST at main's known baseline (the mcp_gateway/search findings that depend on
which optional extras are installed, and the 6 pre-existing mcp_gateway test
failures).
🤖 Generated with Claude Code
https://claude.ai/code/session_015dYktt85Ltj3oK9gG5VBww