chore: bump `pocket-ic` to `release-2026-04-30_04-26-base` by ggreif · Pull Request #6088 · caffeinelabs/motoko

ggreif · 2026-05-06T00:20:14Z

Summary

Bumps pocket-ic-server (and the matching pocket-ic Rust crate) from release-2026-04-16_04-20-base to release-2026-04-30_04-26-base.
Drives the upgrade through the bump-pocket-ic skill — also relocated from .claude/skills/ to .agents/skills/ (matching the bump-rust-nightly precedent in chore: bump rustc-nightly to 2026-05-04 (May the fourth) #6079) and tweaked based on what we learned this run.

Skill retro (folded back into the SKILL.md)

SHA256SUMS is no longer published by recent release-*-base tags (URL 404s). Option A demoted to an optional cross-check; nix-prefetch-url is now the default. Nix rejects tampered downloads on its own, so the cross-check is belt-and-suspenders.
The skill's nix-prefetch-url --print-path | head -1 recipe was unnecessary; plain nix-prefetch-url <url> returns the base32 hash on stdout.
cargo update must run with plain cargo, not nix develop --command cargo … — the dev shell is offline-sandboxed and fails with --offline.
Scope the lockfile change with cargo update -p pocket-ic.
Across consecutive base tags the crate version normally stays the same (here: 13.0.0 → 13.0.0); only the git commit hash moves, so the flake outputHashes hash always changes but the key usually doesn't.

Test plan

nix build .#pocket-ic-server
nix build .#test-runner
CI green (full test suite)

🤖 Generated with Claude Code

Same fix as PR #6009 applied to `gc-tests`. The `tests` matrix on `arm64-linux-16` runs `nix-build-uncached --max-jobs auto`, which schedules three parallel ~3 GB derivation builds and OOMs the runner ("Killed" → "The runner has received a shutdown signal"). Use the matrix `include:` pattern to override `max_jobs` per-OS and plumb it through to `test-blueprint` like `gc-tests` already does. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

github-actions · 2026-05-06T00:41:45Z

Comparing from 53943c3 to 41201d2:
The produced WebAssembly code seems to be completely unchanged.
In terms of gas, no changes are observed in 5 tests.
In terms of size, no changes are observed in 5 tests.

This reverts commit 6c766d8b0bf6d28cf2a0fc92c5b2d9d76eed9a05. PR #6088 with a fresh cache-buster (forcing a real cold rebuild of the GC-variant derivations) hit OOM at max-jobs=2: building 'test-drun-compacting-gc.drv'... building 'test-drun-generational-gc.drv'... Killed nix-build-uncached -build-flags "--max-jobs 2" Run: https://github.com/caffeinelabs/motoko/actions/runs/25411132891/job/74532926605 The GC variants are heavier than the regular test-* derivations (each RTS rebuild with different feature flags = ~3 GB), so two in parallel exceed the 16 GB runner. PR #6090's gc-tests at max-jobs=2 only appeared to pass because that branch's cache-busting populated cachix with a different content hash and the GC variants got pulled instead of rebuilt. For \`tests\` (release-systems-go) the constituent derivations are smaller and max-jobs=2 holds — keeping that bump. The matrix is asymmetric: gc-tests=1, tests=2. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Asymmetric tuning (gc-tests=1, tests=2) is too brittle: PR #6090 showed tests-at-2 passing on a particular cache state, but the gc-tests OOM in PR #6088 demonstrated that the 2-vs-1 distinction depends on which derivations happen to rebuild simultaneously. Future test additions could shift the memory profile of release-systems-go and reintroduce OOMs silently. Going back to a uniform max-jobs=1 on arm64-linux-16. Slower but stable. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Same shape as the `bump-rust-nightly` move (commit 93e5b09): promotes the skill from the user-local `.claude/skills/` to the repo-shared `.agents/skills/` location used by `build-and-test`, `migrating-motoko`, etc. Renames `skill.md` → `SKILL.md` for consistency. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Bumps the `pocket-ic-server` binary from `release-2026-04-16_04-20-base` to the 2026-04-30 base release. SHA256 hashes obtained via `nix-prefetch-url`. (The release does not ship a `SHA256SUMS` file: a `curl -sIL` of the URL returns 404, so the skill's recommended Option A is unusable for these releases.) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Pins the `pocket-ic` crate in `test-runner/Cargo.toml` to the same release tag (`release-2026-04-30_04-26-base`) and refreshes `Cargo.lock`. The crate version is unchanged (13.0.0); only the git commit hash of the tag moves. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The `pocket-ic` crate version stayed at 13.0.0, but the underlying git tag now points at a different commit, so `outputHashes` needs the new fixed-output hash. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Findings from the 2026-04-30 bump: 1. `SHA256SUMS` is no longer published by recent `release-*-base` tags (URL 404s). Demoted Option A to an optional cross-check; promoted `nix-prefetch-url` to the default. Nix already rejects tampered downloads, so the cross-check is belt-and-suspenders, not required. 2. The `nix-prefetch-url --print-path | head -1` recipe was overly ornate; plain `nix-prefetch-url <url>` returns the base32 hash on stdout already. 3. `cargo update` must run with plain `cargo`, not via `nix develop --command` — the dev shell is offline-sandboxed and the command fails with `you are in the offline mode (--offline)`. 4. Use `cargo update -p pocket-ic` to scope the change. 5. Across consecutive `release-*-base` tags the crate version normally stays the same; only the git commit hash moves. Reworded step 6 to reflect that "hash always changes, version usually doesn't". Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

## Summary Extends the same `max_jobs=1` workaround used by `gc-tests` in caffeinelabs#6009 to the `tests` matrix on `arm64-linux-16`. `ubuntu-24-large` keeps `max-jobs=auto`. ## Why `tests (arm64-linux-16, release)` and `tests (arm64-linux-16, debug)` have been OOM-killed by the runner whenever `nix-build-uncached --max-jobs auto` schedules three parallel ~3 GB derivation builds. Most recent observation: https://github.com/caffeinelabs/motoko/actions/runs/25409615273/job/74528169567 (initial CI on the `pocket-ic` bump PR caffeinelabs#6088 — failure was unrelated to the bump itself). ## Why max-jobs=1 (not 2) We probed a higher ceiling but it's not stable enough to ship: - PR caffeinelabs#6090 ran `gc-tests` and `tests` at `max-jobs=2` against a cache-busted `test-runner` and **appeared** to pass. - PR caffeinelabs#6088 with a *different* cache-buster (forcing a different content hash than caffeinelabs#6090's, and therefore a real cold rebuild of the GC-variant derivations) hit OOM on `gc-tests (arm64-linux-16, 2)` after ~2.5 minutes: ``` building 'test-drun-compacting-gc.drv'... building 'test-drun-generational-gc.drv'... Killed nix-build-uncached -build-flags "--max-jobs 2" ``` Run: https://github.com/caffeinelabs/motoko/actions/runs/25411132891/job/74532926605 The result depends on which derivations happen to be cachix-cached vs cold-rebuilt at any given moment. An asymmetric `gc-tests=1, tests=2` configuration would also be brittle: a future test addition could shift the memory profile of `release-systems-go` and silently reintroduce OOMs. Going uniform `max-jobs=1` is slower but stable. ## Test plan - [x] CI green - [x] No `tests (arm64-linux-16, *)` OOM on the resulting run 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

ggreif requested a review from a team as a code owner May 6, 2026 00:20

ggreif self-assigned this May 6, 2026

ggreif enabled auto-merge May 6, 2026 00:24

ggreif mentioned this pull request May 6, 2026

chore: serialise tests builds on arm64-linux-16 #6089

Merged

2 tasks

ggreif force-pushed the gabor/bump-pocket-ic branch from 272a555 to 9f27f75 Compare May 6, 2026 00:31

ggreif force-pushed the gabor/bump-pocket-ic branch 2 times, most recently from c77c14a to eefcb56 Compare May 6, 2026 01:11

ggreif force-pushed the gabor/bump-pocket-ic branch 2 times, most recently from fcce493 to c571caf Compare May 6, 2026 01:51

ggreif and others added 5 commits May 6, 2026 07:52

chore: fix pocket-ic git fetch hash in flake.nix

c3efd49

The `pocket-ic` crate version stayed at 13.0.0, but the underlying git tag now points at a different commit, so `outputHashes` needs the new fixed-output hash. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

ggreif force-pushed the gabor/bump-pocket-ic branch from c571caf to 41201d2 Compare May 6, 2026 05:53

ggreif disabled auto-merge May 6, 2026 06:06

alexandru-uta approved these changes May 6, 2026

View reviewed changes

ggreif added this pull request to the merge queue May 6, 2026

Merged via the queue into master with commit dfc0a50 May 6, 2026
20 checks passed

ggreif deleted the gabor/bump-pocket-ic branch May 6, 2026 07:32

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

chore: bump `pocket-ic` to `release-2026-04-30_04-26-base` - #6088

chore: bump `pocket-ic` to `release-2026-04-30_04-26-base`#6088
ggreif merged 6 commits into
masterfrom
gabor/bump-pocket-ic

ggreif commented May 6, 2026

Uh oh!

github-actions Bot commented May 6, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

ggreif commented May 6, 2026

Summary

Skill retro (folded back into the SKILL.md)

Test plan

Uh oh!

github-actions Bot commented May 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented May 6, 2026 •

edited

Loading