Skip to content

chore: bump wails/v3 beta.4 -> beta.5 + @wailsio/runtime to match (Dependabot #2) - #32

Merged
alicoding merged 1 commit into
mainfrom
chore/wails-v3-beta5
Aug 12, 2026
Merged

chore: bump wails/v3 beta.4 -> beta.5 + @wailsio/runtime to match (Dependabot #2)#32
alicoding merged 1 commit into
mainfrom
chore/wails-v3-beta5

Conversation

@alicoding

Copy link
Copy Markdown
Owner

Summary

Findings requested by the goal brief

(a) NSWindow/NSPanel (ADR-0033's Quick Panel focus-steal limitation): No window/panel API changes in beta.5 at all. The only darwin file touched (application_darwin.go) is an activation-policy-gated fix for activateIgnoringOtherApps (background/agent apps no longer steal focus at launch) -- unrelated to the window/panel class Wails creates. ADR-0033's documented NSWindow-only limitation is unchanged -- nothing to follow up on yet, no redesign attempted.

(b) Dock bounce/badge: Confirmed absent from beta.5 by grepping the actual @wailsio/runtime@3.0.0-beta.5 npm tarball for "bounce"/"badge" -- zero matches. Consistent with the prior session's note that this lands in beta.6+, not beta.5.

Also fixed

@wailsio/runtime was pinned to the literal string "latest" in package.json (not a real semver range) -- which is why it had silently drifted to a resolved beta.1 in the lockfile for months despite go.mod being on beta.4 ("latest" only re-resolves when the lockfile doesn't already satisfy it). Repinned to an exact "3.0.0-beta.5", matching the Go side.

Bindings

Regenerated via task common:generate:bindings, both with the pre-existing beta.4 wails3 CLI and again after upgrading the installed CLI to beta.5 -- zero diff either time. No bound method signature or generation-format changes in this release.

Test plan

  • go vet, golangci-lint (0 issues), go test -race -cover (all packages ok)
  • go build (desktop + server tags) -- both green
  • npm run lint/tsc/boundaries/test (231/231)/build -- all green
  • file-loc-limit, root-file-naming, rules-frontmatter -- green
  • Full local Playwright run (55 spec files): 1 apparent failure in layout.spec.ts's build-badge test, root-caused to a sandboxed-git-worktree-specific artifact -- go build's automatic VCS stamping embedded a different branch's commit into BuildInfo.Revision than this worktree's own checked-out HEAD, reproduced deterministically across 3 rebuilds, even though git rev-parse HEAD in the identical shell/cwd is correct. CI checks out a plain non-worktree clone, so this can't occur there -- treating CI as the authoritative gate for this one. Two other local-only hiccups (quick-panel status race, seed-completeness run timing) passed on Playwright's automatic retry, consistent with ordinary local resource contention.
  • CI green through merge (auto-merge enabled) -- watching closely given this is the highest-scrutiny bump

🤖 Generated with Claude Code

https://claude.ai/code/session_01FYwojT8GdUbYSoggbvEFft

…pendabot #2)

Read the full beta.5 release notes/changelog and diffed the actual
commit range (v3.0.0-beta.4...v3.0.0-beta.5) before touching anything,
per this goal's explicit ask for extra scrutiny on this bump. Findings:

- beta.5 is a small, 3-commit bug-fix release: macOS activation-policy
  focus-steal fix (application_darwin.go), two Linux GTK/WebKit fixes
  (linux_cgo*.go, webview_window_linux.go). No other files changed.
- (a) NSWindow/NSPanel: no window/panel API changes at all --
  application_darwin.go's only change is activation-policy-gated
  activateIgnoringOtherApps, unrelated to window/panel class. ADR-0033's
  documented NSWindow-only limitation (settingsservice_panel.go) is
  UNCHANGED by this release -- no non-activating-panel mechanism to
  adopt, nothing to follow up on yet.
- (b) Dock bounce/badge: confirmed absent by grepping the actual
  @wailsio/runtime@3.0.0-beta.5 npm tarball for "bounce"/"badge" --
  zero matches. The prior session's note that beta.6-era runtime
  unlocks dock-bounce is consistent with this: it is not in beta.5.

@wailsio/runtime was pinned to the literal string "latest" in
package.json (not an actual semver range), which is why it had
silently drifted to a resolved beta.1 in package-lock.json for
months despite go.mod being on beta.4 -- "latest" only re-resolves
when the lockfile doesn't already satisfy it, so it never moved on
its own. Fixed to an exact "3.0.0-beta.5" pin, matching the Go side
exactly (PR #6, the Dependabot PR for this exact bump, was already
closed -- this supersedes it properly instead of leaving the drift).

Bindings: regenerated via `task common:generate:bindings` (both with
the pre-existing beta.4 wails3 CLI and after upgrading the installed
CLI to match beta.5) -- zero diff either time, confirming no bound
method signature or generation-format changes in this release.

Full local suite green: go vet, golangci-lint (0 issues), go test
(-race -cover, all packages ok), go build (desktop + server tags),
eslint/tsc/boundaries/vitest (231/231)/frontend build, file-loc-limit,
root-file-naming, rules-frontmatter.

Local full e2e run (55 spec files): 1 apparent failure in
layout.spec.ts's build-badge test, root-caused to an environment
artifact specific to this sandboxed git-worktree checkout, not a
real regression -- go build's automatic VCS stamping
(-buildvcs=auto) embedded a DIFFERENT branch's commit
(docs/ratified-queue-reorder's tip) into the e2e server binary's
BuildInfo.Revision, rather than this worktree's own checked-out
HEAD, even though `git rev-parse HEAD` in the exact same shell/cwd
correctly reports this branch's commit -- reproduced deterministically
across three separate rebuilds. CI checks out a plain, non-worktree
clone, so this class of mismatch cannot occur there; treating CI's
run (this branch's actual DoD gate) as authoritative rather than
chasing a local-sandbox-only artifact. Two other local-only test
hiccups (quick-panel status-text race, seed-completeness run-timing)
passed on Playwright's own automatic retry, consistent with ordinary
4-worker local resource contention rather than anything wails-related.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FYwojT8GdUbYSoggbvEFft
@alicoding
alicoding enabled auto-merge (squash) August 12, 2026 20:31
@alicoding
alicoding merged commit 255d462 into main Aug 12, 2026
29 of 31 checks passed
alicoding added a commit that referenced this pull request Aug 15, 2026
…pendabot #2) (#32)

Read the full beta.5 release notes/changelog and diffed the actual
commit range (v3.0.0-beta.4...v3.0.0-beta.5) before touching anything,
per this goal's explicit ask for extra scrutiny on this bump. Findings:

- beta.5 is a small, 3-commit bug-fix release: macOS activation-policy
  focus-steal fix (application_darwin.go), two Linux GTK/WebKit fixes
  (linux_cgo*.go, webview_window_linux.go). No other files changed.
- (a) NSWindow/NSPanel: no window/panel API changes at all --
  application_darwin.go's only change is activation-policy-gated
  activateIgnoringOtherApps, unrelated to window/panel class. ADR-0033's
  documented NSWindow-only limitation (settingsservice_panel.go) is
  UNCHANGED by this release -- no non-activating-panel mechanism to
  adopt, nothing to follow up on yet.
- (b) Dock bounce/badge: confirmed absent by grepping the actual
  @wailsio/runtime@3.0.0-beta.5 npm tarball for "bounce"/"badge" --
  zero matches. The prior session's note that beta.6-era runtime
  unlocks dock-bounce is consistent with this: it is not in beta.5.

@wailsio/runtime was pinned to the literal string "latest" in
package.json (not an actual semver range), which is why it had
silently drifted to a resolved beta.1 in package-lock.json for
months despite go.mod being on beta.4 -- "latest" only re-resolves
when the lockfile doesn't already satisfy it, so it never moved on
its own. Fixed to an exact "3.0.0-beta.5" pin, matching the Go side
exactly (PR #6, the Dependabot PR for this exact bump, was already
closed -- this supersedes it properly instead of leaving the drift).

Bindings: regenerated via `task common:generate:bindings` (both with
the pre-existing beta.4 wails3 CLI and after upgrading the installed
CLI to match beta.5) -- zero diff either time, confirming no bound
method signature or generation-format changes in this release.

Full local suite green: go vet, golangci-lint (0 issues), go test
(-race -cover, all packages ok), go build (desktop + server tags),
eslint/tsc/boundaries/vitest (231/231)/frontend build, file-loc-limit,
root-file-naming, rules-frontmatter.

Local full e2e run (55 spec files): 1 apparent failure in
layout.spec.ts's build-badge test, root-caused to an environment
artifact specific to this sandboxed git-worktree checkout, not a
real regression -- go build's automatic VCS stamping
(-buildvcs=auto) embedded a DIFFERENT branch's commit
(docs/ratified-queue-reorder's tip) into the e2e server binary's
BuildInfo.Revision, rather than this worktree's own checked-out
HEAD, even though `git rev-parse HEAD` in the exact same shell/cwd
correctly reports this branch's commit -- reproduced deterministically
across three separate rebuilds. CI checks out a plain, non-worktree
clone, so this class of mismatch cannot occur there; treating CI's
run (this branch's actual DoD gate) as authoritative rather than
chasing a local-sandbox-only artifact. Two other local-only test
hiccups (quick-panel status-text race, seed-completeness run-timing)
passed on Playwright's own automatic retry, consistent with ordinary
4-worker local resource contention rather than anything wails-related.


Claude-Session: https://claude.ai/code/session_01FYwojT8GdUbYSoggbvEFft

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant