release: fix Homebrew/Scoop publish auth (checkout+token, not URL-embedded PAT) by MelbourneDeveloper · Pull Request #60 · Nimblesite/Basilisk

MelbourneDeveloper · 2026-05-30T05:25:13Z

Pull Request

TLDR

Fix the broken Homebrew/Scoop tap publishing in the release pipeline (the actual blocker), and along the way scrub the VS Code sidebar + docs of features that were advertised but never wired up.

What Was Added

vscode-extension/src/test/suite/info-panel.test.ts (new, +121) — VSIX E2E test that renders InfoPanelProvider, pins the exact set of Feature Status toggles that actually do something, and proves the uv Integration toggle shows/hides the uv Quick Actions. Fails if a no-op toggle is ever re-added.
docs/specs/CHECKER-STUB-RESOLUTION-SPEC.md (new spec) — documents .pyi stub discovery/resolution and the stub-suggestion behaviour.
docs/plans/EXTENSION-ACTIVITY-PANEL-PLAN.md → #EXTACT-PLAN-FEATURE-TOGGLES — a step-by-step plan for making each removed toggle real (server config struct, per-feature gating, and the specific VSIX test each one needs before it may return).
Release pipeline jobs split out cleanly: build → release → vsix → release-assets → publish-vsix → deploy-pages → publish-homebrew / publish-scoop, with the GitHub Release archives as the single source of truth for downstream jobs (preserves the Unix exec bit; raw Actions artifacts strip it).

What Was Changed / Deleted

.github/workflows/release.yml (the headline fix): replaced the rejected git clone https://x-access-token:${BREW_SCOOP_PAT}@github.com/... tap auth with actions/checkout@v4 + token: ${{ secrets.BREW_SCOOP_PAT }} (checkout configures the PAT as an HTTP extraheader, which GitHub accepts; the URL-embedded PAT fails with "Password authentication is not supported"). The VSIX job now downloads the per-platform binary from the published release instead of recompiling — binaries are no longer built twice.
vscode-extension/src/info-panel.ts: removed 6 no-op Feature Status toggles (Inlay Hints Params/Types, Ruff Integration, Test Explorer, Debugger, AI Typing). Root cause: the LSP server's did_change_configuration only parses analysisMode + testExplorer, so every other forwarded setting was silently dropped — those toggles wrote settings nothing read. Kept only the two with a real, namesake effect: Type Checking (gates diagnostic publication) and uv Integration (gates the uv Quick Actions + Server Info rows).
vscode-extension/package.json: deleted the dead aiTyping.enabled / aiTyping.provider settings; relabelled inlayHints.* and ruff.* descriptions as "Reserved — not yet read by the server".
vscode-extension/README.md: corrected the Basilisk Info panel feature list and the settings table to match reality.
Website (website/src/**, en + zh): scrubbed the docs of the same lies — configuration.md now marks inlayHints/ruff/testExplorer settings "reserved"; installation.md replaces the fake inlayHints/ruff Zed config example with the genuinely-honored analysisMode. Plus a broader docs/landing refresh (comparison, migration, index.njk, blog, favicon/logo/og-image).
CLAUDE.md / docs/INDEX.md / crates/basilisk-config/README.md: project-instruction and index housekeeping; lspkit migration notes.

How Tests Prove It Works

Release pipeline (the real proof): tag v0.4.3 was cut from this branch and the Release workflow ran green end-to-end — all 16 jobs success, including Publish Homebrew formula and Publish Scoop manifest (the exact jobs this branch fixes) and Publish VSIX to the Marketplace. Run: Nimblesite/Basilisk actions run 26680555278.
Sidebar: make _test_vsix → 305 passing / 0 failing, VSIX coverage 85% ≥ 84% threshold. New Basilisk Info Panel Contents suite:
- Feature Status lists exactly the toggles with a real effect
- Feature Status hides every no-op toggle
- uv Integration shows uv Quick Actions when enabled
- uv Integration hides uv Quick Actions when disabled
Lint: make lint green (clippy --workspace --all-targets -D warnings + eslint); cargo fmt --check and ruff format --check clean.

Spec / Doc Changes

EXTENSION-ACTIVITY-PANEL-SPEC.md#EXTACT-INFO-FEATURE-STATUS — rewritten: shipped vs. not-yet-implemented toggles, with the rule that a toggle may only exist if it has a real effect proven by a VSIX test.
EXTENSION-ACTIVITY-PANEL-PLAN.md#EXTACT-PLAN-FEATURE-TOGGLES — new implementation plan for the removed toggles.
CHECKER-STUB-RESOLUTION-SPEC.md — new spec.
CHECKER-ARCHITECTURE-SPEC.md, LSP-PROFILER-PROCESS-PANEL-PLAN.md — housekeeping.

Breaking Changes

No. The removed aiTyping.* settings and the 6 retired toggles never had any effect, so removing them changes no observable behaviour. inlayHints.* / ruff.* settings remain declared (relabelled "reserved") for forward compatibility.

… suite + website; docs/site refresh (#59) ## TLDR Fixes the new-module import false positive (#53) with live workspace re-resolution, stops ANSI escape garbage in piped CLI/LSP logs, simplifies Shipwright binary resolution to user-setting + bundled, and adds a benchmark fixture suite plus a website benchmark section — alongside a docs/site refresh. ## What Was Added? - **`crates/basilisk-lsp/tests/lsp/ws_test_cross_module.rs`** (new, 10 e2e tests) covering cross-module analysis under `[ANALYSIS-CROSSLSP]`, including two regressions for #53: `new_file_resolves_first_party_sibling_import` and `created_module_clears_unresolved_import_in_dependents`. - **`WorkspaceIndex::forget_file`** plus **`recheck_all_files` / `reresolve_imports_and_recheck`** in `workspace.rs` (`[ANALYSIS-INCR-IMPORTS]`) — drop deleted entries and re-check every indexed file against current resolution. - **Tracing ANSI regression test** `tracing_logs_emit_no_ansi_on_piped_stderr` in `crates/basilisk-cli/tests/cli_binary_tests.rs`. - **Benchmark fixtures**: `benchmarks/fixtures/e0010_unresolved_import.py`, `e0012_argument_type_mismatch.py`, `e0014_assignment_incompatibility.py` (2k-line stress fixtures), reworked `benchmarks/run.sh`, and `benchmarks/status/darwin-arm64-apple-m4-max.csv`. - **Website benchmark section**: `website/src/_data/benchmarks.js` + `website/src/_includes/benchmark-section.njk` rendering benchmark results on the site. - **`scripts/gen_rules_reference.py`** — generates the rules reference docs. - **`docs/plans/ROADMAP-NEXT-STEPS-PLAN.md`** (new plan). ## What Was Changed or Deleted? - **#53 fix — `crates/basilisk-lsp/src/server/document.rs`**: the file-watcher now distinguishes `CREATED` from `CHANGED`. A newly-created module re-resolves the whole workspace and re-publishes so dependents' stale `BSK-E0010 "cannot resolve import"` clears without an LSP reload; a pure content change keeps the cheaper per-file reload. Deleted files are forgotten first so a recheck cannot resurrect their diagnostics. `[ANALYSIS-INCR-IMPORTS]`. - **CLI logs — `crates/basilisk-cli/src/main.rs`**: gate ANSI colour on `std::io::IsTerminal::is_terminal(stderr)`, so the LSP subprocess (piped stderr) emits clean text instead of raw escape sequences in the editor output channel. - **Shipwright — `shipwright.json`**: binary resolution sources narrowed to `["user-setting", "bundled"]` (env / pkgmgr / cargo-bin / path / lsp-initialize removed). The VSIX ships and resolves its bundled binary by default; a user override is still honoured. - **VS Code test helpers — `debug-integration.test.ts`, `test-helpers.ts`**: drop ad-hoc PATH/platform binary discovery in favour of the Shipwright-staged bundled binary. - **Docs/site refresh**: `Claude.md` → `CLAUDE.md`, expanded `CHECKER-ARCHITECTURE-SPEC.md`, `LSP-ARCHITECTURE-SPEC.md`, `VSIX-SPEC.md`, `README.md`, and a large en/zh website docs overhaul (comparison, configuration, rules/*, removed standalone mojo-safety rule pages). - **Build/CI**: `Makefile`, `.github/workflows/{ci,release,deploy-pages}.yml`, `.gitignore`, `coverage-thresholds.json` (metadata), `.claude/skills/*` metadata, removed the `tmc` skill. ## How Do The Automated Tests Prove It Works? - **#53**: `new_file_resolves_first_party_sibling_import` creates a new module *after* the workspace scan and asserts its first-party sibling import resolves like a pre-existing file; `created_module_clears_unresolved_import_in_dependents` asserts a dependent's `BSK-E0010` is re-published as cleared once the module appears — both pass under `./scripts/test-rust.sh`. - **Logs**: `tracing_logs_emit_no_ansi_on_piped_stderr` runs `basilisk check` with `BASILISK_LOG=info` over captured (piped) stderr and asserts zero ANSI escapes. - **Shipwright/VSIX**: VS Code E2E "Shipwright resolves basilisk from the bundled VSIX binary by default" plus the full 301-test suite pass (85% ≥ 84% coverage); `verify-shipwright.mjs manifest` and `versions` both pass. - **Full local gate run** (mirrors CI): Rust coverage thresholds met on all 10 crates; Zed (96 tests), Neovim (39% ≥ 30%), and mutation `working` baseline (90.54%, no regression) all green. ## Spec / Doc Changes - `[ANALYSIS-INCR-IMPORTS]` / `[ANALYSIS-CROSSLSP]` documented in the LSP/checker specs; `CHECKER-ARCHITECTURE-SPEC.md` and `LSP-ARCHITECTURE-SPEC.md` updated; `Claude.md` renamed to `CLAUDE.md`; new `ROADMAP-NEXT-STEPS-PLAN.md`. ## Breaking Changes - Shipwright binary resolution now considers only `user-setting` and `bundled` sources (previously also env / package managers / `cargo-bin` / PATH / LSP-initialize). Environments that relied on PATH or `BASILISK_EXECUTABLE_PATH` discovery should set the user setting or use the bundled binary. Fixes #53 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Co-authored-by: Abdul Shakoor <60779242+abdushakoor12@users.noreply.github.com> Co-authored-by: abdushakoor12 <abdushakoor009@gmail.com>

… suite + website; docs/site refresh (#59) ## TLDR Fixes the new-module import false positive (#53) with live workspace re-resolution, stops ANSI escape garbage in piped CLI/LSP logs, simplifies Shipwright binary resolution to user-setting + bundled, and adds a benchmark fixture suite plus a website benchmark section — alongside a docs/site refresh. ## What Was Added? - **`crates/basilisk-lsp/tests/lsp/ws_test_cross_module.rs`** (new, 10 e2e tests) covering cross-module analysis under `[ANALYSIS-CROSSLSP]`, including two regressions for #53: `new_file_resolves_first_party_sibling_import` and `created_module_clears_unresolved_import_in_dependents`. - **`WorkspaceIndex::forget_file`** plus **`recheck_all_files` / `reresolve_imports_and_recheck`** in `workspace.rs` (`[ANALYSIS-INCR-IMPORTS]`) — drop deleted entries and re-check every indexed file against current resolution. - **Tracing ANSI regression test** `tracing_logs_emit_no_ansi_on_piped_stderr` in `crates/basilisk-cli/tests/cli_binary_tests.rs`. - **Benchmark fixtures**: `benchmarks/fixtures/e0010_unresolved_import.py`, `e0012_argument_type_mismatch.py`, `e0014_assignment_incompatibility.py` (2k-line stress fixtures), reworked `benchmarks/run.sh`, and `benchmarks/status/darwin-arm64-apple-m4-max.csv`. - **Website benchmark section**: `website/src/_data/benchmarks.js` + `website/src/_includes/benchmark-section.njk` rendering benchmark results on the site. - **`scripts/gen_rules_reference.py`** — generates the rules reference docs. - **`docs/plans/ROADMAP-NEXT-STEPS-PLAN.md`** (new plan). ## What Was Changed or Deleted? - **#53 fix — `crates/basilisk-lsp/src/server/document.rs`**: the file-watcher now distinguishes `CREATED` from `CHANGED`. A newly-created module re-resolves the whole workspace and re-publishes so dependents' stale `BSK-E0010 "cannot resolve import"` clears without an LSP reload; a pure content change keeps the cheaper per-file reload. Deleted files are forgotten first so a recheck cannot resurrect their diagnostics. `[ANALYSIS-INCR-IMPORTS]`. - **CLI logs — `crates/basilisk-cli/src/main.rs`**: gate ANSI colour on `std::io::IsTerminal::is_terminal(stderr)`, so the LSP subprocess (piped stderr) emits clean text instead of raw escape sequences in the editor output channel. - **Shipwright — `shipwright.json`**: binary resolution sources narrowed to `["user-setting", "bundled"]` (env / pkgmgr / cargo-bin / path / lsp-initialize removed). The VSIX ships and resolves its bundled binary by default; a user override is still honoured. - **VS Code test helpers — `debug-integration.test.ts`, `test-helpers.ts`**: drop ad-hoc PATH/platform binary discovery in favour of the Shipwright-staged bundled binary. - **Docs/site refresh**: `Claude.md` → `CLAUDE.md`, expanded `CHECKER-ARCHITECTURE-SPEC.md`, `LSP-ARCHITECTURE-SPEC.md`, `VSIX-SPEC.md`, `README.md`, and a large en/zh website docs overhaul (comparison, configuration, rules/*, removed standalone mojo-safety rule pages). - **Build/CI**: `Makefile`, `.github/workflows/{ci,release,deploy-pages}.yml`, `.gitignore`, `coverage-thresholds.json` (metadata), `.claude/skills/*` metadata, removed the `tmc` skill. ## How Do The Automated Tests Prove It Works? - **#53**: `new_file_resolves_first_party_sibling_import` creates a new module *after* the workspace scan and asserts its first-party sibling import resolves like a pre-existing file; `created_module_clears_unresolved_import_in_dependents` asserts a dependent's `BSK-E0010` is re-published as cleared once the module appears — both pass under `./scripts/test-rust.sh`. - **Logs**: `tracing_logs_emit_no_ansi_on_piped_stderr` runs `basilisk check` with `BASILISK_LOG=info` over captured (piped) stderr and asserts zero ANSI escapes. - **Shipwright/VSIX**: VS Code E2E "Shipwright resolves basilisk from the bundled VSIX binary by default" plus the full 301-test suite pass (85% ≥ 84% coverage); `verify-shipwright.mjs manifest` and `versions` both pass. - **Full local gate run** (mirrors CI): Rust coverage thresholds met on all 10 crates; Zed (96 tests), Neovim (39% ≥ 30%), and mutation `working` baseline (90.54%, no regression) all green. ## Spec / Doc Changes - `[ANALYSIS-INCR-IMPORTS]` / `[ANALYSIS-CROSSLSP]` documented in the LSP/checker specs; `CHECKER-ARCHITECTURE-SPEC.md` and `LSP-ARCHITECTURE-SPEC.md` updated; `Claude.md` renamed to `CLAUDE.md`; new `ROADMAP-NEXT-STEPS-PLAN.md`. ## Breaking Changes - Shipwright binary resolution now considers only `user-setting` and `bundled` sources (previously also env / package managers / `cargo-bin` / PATH / LSP-initialize). Environments that relied on PATH or `BASILISK_EXECUTABLE_PATH` discovery should set the user setting or use the bundled binary. Fixes #53 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: abdushakoor12 <abdushakoor009@gmail.com>

…edded PAT) The publish-homebrew/publish-scoop jobs cloned the tap/bucket via `git clone https://x-access-token:$BREW_SCOOP_PAT@github.com/...`, which GitHub rejects even with a valid PAT: remote: Invalid username or token. Password authentication is not supported fatal: Authentication failed Switch to the pattern proven in Nimblesite/dart_mutant: check out the target repo with actions/checkout (repository + token + path), render into it, then commit & push using checkout's configured credentials. Drops the now-redundant empty-PAT guard and BREW_SCOOP_PAT env on the render steps. Refs Nimblesite/Shipwright#2. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…unners All CI jobs targeted the custom 'ubuntu-22.04-8-cores' larger-runner label. After the repo moved from MelbourneDeveloper to the Nimblesite org, those org-scoped larger runners are not provisioned, so every CI job sat queued indefinitely waiting for a runner that never came (the release workflow was unaffected because it uses GitHub-hosted ubuntu-latest/macos-15/windows-latest). Switch all 9 jobs to the standard ubuntu-22.04 hosted runner so CI can actually schedule.

…ld scripts run Switching CI to ubuntu-22.04 fixed the queue hang but introduced a GLIBC mismatch: the Zed job's cached serde build-script binary and the cargo-mutants release downloaded by taiki-e/install-action are both linked against GLIBC_2.39, which ubuntu-22.04 (GLIBC 2.35) cannot load ('version GLIBC_2.39 not found'). ubuntu-24.04 ships GLIBC 2.39 and is a standard GitHub-hosted runner (the release workflow already runs green on ubuntu-latest=24.04), so it both schedules immediately and satisfies the toolchain ABI.

The Zed job kept failing even on ubuntu-24.04: Swatinem/rust-cache restored a host build-script binary (serde's build-script-build) linked against a glibc the runner does not provide ('GLIBC_2.39 not found'), and cache-on-failure:true kept re-saving the poisoned artifact. rust-cache keys per job, so only Zed was affected (test-rust and the others passed). Bump the Zed job's prefix-key to force a clean rebuild from source against the runner's own glibc.

After moving CI off the retired 8-core larger runners to the standard ubuntu-24.04 hosted runner, the mutation-testing shards began failing non-deterministically — a different shard timed out on each run (shard 1, then shards 2 and 4), which is the signature of a wall-clock timeout rather than a surviving mutant (a real missed mutant is deterministic and fails the same shard every run). cargo-mutants rebuilds and reruns the whole test suite once per mutant and is heavily CPU-bound; the 20-minute budget was calibrated for 8 cores. Raise the per-shard timeout to 60 minutes for comfortable headroom; the gate itself is unchanged (the merge job still enforces the mutation score).

…edded PAT) (#60) # Pull Request ## TLDR Fix the broken Homebrew/Scoop tap publishing in the release pipeline (the actual blocker), and along the way scrub the VS Code sidebar + docs of features that were advertised but never wired up. ## What Was Added - **`vscode-extension/src/test/suite/info-panel.test.ts`** (new, +121) — VSIX E2E test that renders `InfoPanelProvider`, pins the exact set of Feature Status toggles that actually do something, and proves the uv Integration toggle shows/hides the uv Quick Actions. Fails if a no-op toggle is ever re-added. - **`docs/specs/CHECKER-STUB-RESOLUTION-SPEC.md`** (new spec) — documents `.pyi` stub discovery/resolution and the stub-suggestion behaviour. - **`docs/plans/EXTENSION-ACTIVITY-PANEL-PLAN.md` → `#EXTACT-PLAN-FEATURE-TOGGLES`** — a step-by-step plan for making each removed toggle *real* (server config struct, per-feature gating, and the specific VSIX test each one needs before it may return). - Release pipeline jobs split out cleanly: `build` → `release` → `vsix` → `release-assets` → `publish-vsix` → `deploy-pages` → `publish-homebrew` / `publish-scoop`, with the GitHub Release archives as the single source of truth for downstream jobs (preserves the Unix exec bit; raw Actions artifacts strip it). ## What Was Changed / Deleted - **`.github/workflows/release.yml` (the headline fix):** replaced the rejected `git clone https://x-access-token:${BREW_SCOOP_PAT}@github.com/...` tap auth with `actions/checkout@v4` + `token: ${{ secrets.BREW_SCOOP_PAT }}` (checkout configures the PAT as an HTTP extraheader, which GitHub accepts; the URL-embedded PAT fails with "Password authentication is not supported"). The VSIX job now downloads the per-platform binary from the published release instead of recompiling — binaries are no longer built twice. - **`vscode-extension/src/info-panel.ts`:** removed 6 no-op Feature Status toggles (Inlay Hints Params/Types, Ruff Integration, Test Explorer, Debugger, AI Typing). Root cause: the LSP server's `did_change_configuration` only parses `analysisMode` + `testExplorer`, so every other forwarded setting was silently dropped — those toggles wrote settings nothing read. Kept only the two with a real, namesake effect: **Type Checking** (gates diagnostic publication) and **uv Integration** (gates the uv Quick Actions + Server Info rows). - **`vscode-extension/package.json`:** deleted the dead `aiTyping.enabled` / `aiTyping.provider` settings; relabelled `inlayHints.*` and `ruff.*` descriptions as "Reserved — not yet read by the server". - **`vscode-extension/README.md`:** corrected the Basilisk Info panel feature list and the settings table to match reality. - **Website (`website/src/**`, en + zh):** scrubbed the docs of the same lies — `configuration.md` now marks `inlayHints`/`ruff`/`testExplorer` settings "reserved"; `installation.md` replaces the fake `inlayHints`/`ruff` Zed config example with the genuinely-honored `analysisMode`. Plus a broader docs/landing refresh (comparison, migration, index.njk, blog, favicon/logo/og-image). - **`CLAUDE.md` / `docs/INDEX.md` / `crates/basilisk-config/README.md`:** project-instruction and index housekeeping; `lspkit` migration notes. ## How Tests Prove It Works - **Release pipeline (the real proof):** tag **v0.4.3** was cut from this branch and the Release workflow ran green end-to-end — **all 16 jobs success**, including `Publish Homebrew formula` and `Publish Scoop manifest` (the exact jobs this branch fixes) and `Publish VSIX` to the Marketplace. Run: `Nimblesite/Basilisk` actions run `26680555278`. - **Sidebar:** `make _test_vsix` → **305 passing / 0 failing**, VSIX coverage **85% ≥ 84%** threshold. New `Basilisk Info Panel Contents` suite: - `Feature Status lists exactly the toggles with a real effect` - `Feature Status hides every no-op toggle` - `uv Integration shows uv Quick Actions when enabled` - `uv Integration hides uv Quick Actions when disabled` - **Lint:** `make lint` green (clippy `--workspace --all-targets -D warnings` + eslint); `cargo fmt --check` and `ruff format --check` clean. ## Spec / Doc Changes - `EXTENSION-ACTIVITY-PANEL-SPEC.md#EXTACT-INFO-FEATURE-STATUS` — rewritten: shipped vs. not-yet-implemented toggles, with the rule that a toggle may only exist if it has a real effect proven by a VSIX test. - `EXTENSION-ACTIVITY-PANEL-PLAN.md#EXTACT-PLAN-FEATURE-TOGGLES` — new implementation plan for the removed toggles. - `CHECKER-STUB-RESOLUTION-SPEC.md` — new spec. - `CHECKER-ARCHITECTURE-SPEC.md`, `LSP-PROFILER-PROCESS-PANEL-PLAN.md` — housekeeping. ## Breaking Changes No. The removed `aiTyping.*` settings and the 6 retired toggles never had any effect, so removing them changes no observable behaviour. `inlayHints.*` / `ruff.*` settings remain declared (relabelled "reserved") for forward compatibility. --------- Co-authored-by: Abdul Shakoor <60779242+abdushakoor12@users.noreply.github.com> Co-authored-by: abdushakoor12 <abdushakoor009@gmail.com>

MelbourneDeveloper and others added 5 commits May 30, 2026 14:30

Merge branch 'main' into fixrelease

0dd7ba7

fix

22f670d

MelbourneDeveloper closed this May 30, 2026

MelbourneDeveloper added 8 commits May 30, 2026 15:27

Fixes

187a463

fixes

95acfbf

Merge branch 'fixrelease' into fix/release-tap-auth

50c5107

Fixes

533b9c5

Don't build the binaries twice

1885669

Doco fixes

92e3b39

Fix release process

afa6c57

Fixes

f03058e

MelbourneDeveloper reopened this May 30, 2026

MelbourneDeveloper added 2 commits May 30, 2026 20:18

ci: retrigger pipeline for PR #60

605a3e7

MelbourneDeveloper closed this May 30, 2026

MelbourneDeveloper reopened this May 30, 2026

MelbourneDeveloper added 2 commits May 30, 2026 20:41

MelbourneDeveloper merged commit 8a054e3 into main May 30, 2026
12 checks passed

MelbourneDeveloper deleted the fix/release-tap-auth branch May 30, 2026 19:35

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

release: fix Homebrew/Scoop publish auth (checkout+token, not URL-embedded PAT)#60

release: fix Homebrew/Scoop publish auth (checkout+token, not URL-embedded PAT)#60
MelbourneDeveloper merged 18 commits into
mainfrom
fix/release-tap-auth

MelbourneDeveloper commented May 30, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

MelbourneDeveloper commented May 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request

TLDR

What Was Added

What Was Changed / Deleted

How Tests Prove It Works

Spec / Doc Changes

Breaking Changes

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

MelbourneDeveloper commented May 30, 2026 •

edited

Loading