test(compiler): pin that a file edit moves no server-function id

[frenzzy]

Server-function ids are the addresses deployed clients already hold, so keying them on identity rather than position (7c551ae, #3109) is a wire contract — and nothing currently guards it.

The fixture suite cannot: every fixture in __tests__/directives/ is a single fixed source with frozen expected output, so it pins the id format and has no second version to compare against. The positional scheme passed an identically shaped suite; the fix regenerated 137 files. Reintroducing position goes green after exactly the regeneration a developer performs reflexively when snapshots fail — and the failure it lets through is the silent one: an old tab calling getUser reaching deleteUser with a 200.

This is the differential half. Compile two versions of one file and require every surviving name to keep its address:

• a function appended at the end
• a function inserted at the top
• the first function deleted
• the two reordered
• a body edited

Plus the two consequences worth stating rather than discovering:

• a rename is a new address — the old id belongs to nothing, so an old client gets a clean 404 (A stale server-function id is an undistinguishable 404, so version skew cannot be recovered #3110) instead of another function;
• among functions sharing a descriptive name, position still decides (<name>-<hash>-<n>), which no fixture covers today.

Checked by mutation

Reverting create_id to format!("{}-{}", self.hash, self.count - 1) fails 4 of the 7:

expected 'c8cb6025-1' to be 'c8cb6025-0'
expected 'c8cb6025-0' to be 'c8cb6025-1'
expected [ 'c8cb6025-0', 'c8cb6025-1' ] to not include 'c8cb6025-1'
expected 'c8cb6025-0' to match /^handler-[0-9a-f]{1,8}$/

The other three pass under both schemes — inserting at the top, deleting the first function, and editing a body never moved an id even positionally. Worth keeping for exactly that reason: most edits look safe, which is why the bug survived. Deleting the first function is the surprising member of that set, and it is surprising only because the old numbering ran in reverse.

Compiler suite green: 29 files, 4325 passed. Tests only, no changeset.

Raised as a gap in #3120.

[changeset-bot]

⚠️ No Changeset found

Latest commit: 6194ee7

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[codspeed-hq]

Merging this PR will not alter performance

✅ 136 untouched benchmarks
⏩ 132 skipped benchmarks¹

---

_{Comparing frenzzy:test/compiler-id-stability (6194ee7) with next (5230666)}

Footnotes

1. 132 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[frenzzy]

Superseded by d6a4a52 — closing.

Your directives-id-scheme.test.js covers this better than mine did. Both were rebuilt against a create_id reverted to <hash>-<ordinal>:

directives-id-scheme.test.js     6 failed | 1 passed
directives-id-stability.test.js  4 failed | 3 passed

Deriving the ids from an independent implementation catches a change to the hash, the path derivation or the format, where my differential version only notices when an edit happens to move something — and three of my cases pass under both schemes, since inserting at the top, deleting the first function and editing a body never moved an id even positionally. That is a weaker signal for the same regression.

The README line is the part I would not have thought to write, and it is probably the most durable of the three: a fixture regeneration that changes an id is a protocol change, not churn. That is the sentence that stops the next person waving it through.