* refactor: make the app policy's principal the authority for its identity
* fix: align the app backfill with the sibling migration and audit the uncached address
* chore: refresh the sqlx cache after rebasing onto the merged base
* fix: resolve the app execution address uncached, it decides the job's authorization
* chore: cache the EE queries at the ref this branch pins
* chore: cache the EE queries at the ref this branch pins
* fix: derive the app draft's on-behalf-of address on read
* chore: cache the query the draft derivation test added
* fix: derive the app identity on the draft-table and version reads too
* docs: state the draft resolver's authorization contract
* fix: resolve a draft's principal against workspace membership only
* chore: cache the membership lookup the draft resolver added
* fix: drop an unresolvable draft's address instead of leaving it stale
* perf: evict the address cache on change so app dispatch can read it
* fix: evict on superadmin role changes, not only address changes
* refactor: make the app policy's address optional instead of derived on read
* fix: follow an external superadmin's rename into the apps that name them
* docs: state the removal gate once, and correctly
* refactor: drop the app-policy version constant that gated nothing
* docs: drop the last reference to the removed constant
* perf: read the address cache everywhere now that eviction reaches every replica
* fix: keep persisted addresses off the cache the poller evicts asynchronously
* docs: state where the cached address is accepted and where it is not
* docs: keep the cache rule in one place and drop the stale premise
* docs: sort the two lookups by how long a wrong answer lives
* fix: resolve the schedule address uncached where it is written to the row
* docs: name the release this actually ships in
* perf: evict a superadmin's key per workspace instead of the whole cache
* fix: evict every alias a superadmin principal can be spelled as
* docs: describe the trigger as it is
* docs: cover the round-tripped read in the cache rule
* docs: record why a stale dispatch address cannot escalate
* fix: validate a dispatch address against the principal's live binding
* fix: carry the validated address through to the job row and token
* fix: record the validated address on the job row, not the one handed in
* test: run the substep tag check as the non-superadmin it means to test
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* fix: rewrite a stored app address that disagrees with its principal
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* docs: record the accepted staleness window of the cached dispatch address
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* fix: record the validated address on the job's audit row
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* docs: record the accepted rename race of pre-transaction identity resolution
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* docs: separate the app's stored address from the derived one in the resolver doc
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* docs: describe the job identity fast path the push comments skipped
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* fix: backfill a legacy group-prefixed username as the group it names
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* fix: resolve a schedule edit's identity before opening its transaction
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* fix: never resolve a disabled member to a same-named superadmin
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* docs: state what the email-change notify buys, and rewrap two comment lines
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* fix: keep a group's runnables when offboarding a legacy group-prefixed member
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* fix: read the app author from the stored address, as execution does
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* docs: record the rename race's full consequence as a known, accepted limitation
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
* docs: record the keep-target group address case as a known, accepted limitation
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JY4bBCR1q2c5XB8s2r7Ysc
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: show when the last git auto-pull status was recorded
* chore: bump ee-repo-ref for the auto-pull status fix
* fix: show the git auto-pull status age with TimeAgo instead of a year-less date
* test: pin that a stale auto-pull recovery cannot overwrite a newer state
* chore: bump ee-repo-ref for the conditional auto-pull recovery
* fix: keep TimeAgo counting past the first hour in noSeconds mode
* chore: bump ee-repo-ref for the clear_auto_pull_failure contract note
* fix: guard TimeAgo's boundary scheduler against invalid dates and pin same-head newer failures
* chore: bump ee-repo-ref for the timestamp-guarded auto-pull recovery
* test: cover a same-second newer failure surviving a stale auto-pull recovery
* chore: bump ee-repo-ref for the whole-failure recovery match
* test: name the recovery helper after its input, not its staleness
* chore: update ee-repo-ref to c6df9fdd9826efb40d3586a9f97d17dee98ac6ef
This commit updates the EE repository reference after PR #793 was merged in windmill-ee-private.
Previous ee-repo-ref: 6aff80b80cae4944a4a78a6b9244019bc37f368b
New ee-repo-ref: c6df9fdd9826efb40d3586a9f97d17dee98ac6ef
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Ruben Fiszel <ruben@windmill.dev>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix: bring back Publish to Hub for scripts
Publishing to the Hub moved to the folder-level flow, which publishes a
whole project and needs a workspace admin. That left no way to share a
single script, which is what private hubs mostly use the Hub for.
Restore the "Publish to Hub" item on the script detail page and in the
script list row menu. Both open the Hub's script submission form prefilled
with the script, on whichever Hub the instance is configured to use, and
are hidden when the instance disables the Hub. Flows and apps still reach
the Hub only inside a project.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: open the Hub tab before fetching, and hide Publish to Hub from operators
The script list row has to fetch the script before it can build the Hub
URL, and Safari refuses window.open after an await, so the tab never
opened there. Claim it inside the click with claimTab(), point it at the
Hub once the script loads, and close it with a toast if the fetch fails.
A blocked popup falls back to a late window.open, and says so if that is
blocked too.
Operators can't write scripts, so the row menu now hides the item from
them, as the script page's menu already does. The script page opens the
Hub with noopener, and scriptToHubUrl takes the script instead of eight
positional arguments.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* [ee] fix: accept any hub version of the git sync script in the token check
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sm7h47G1rC3qYADfkCtTiZ
* chore: update ee-repo-ref to c9b043f2860fdae150c8c4bf03f3ec98b7f300e5
This commit updates the EE repository reference after PR #792 was merged in windmill-ee-private.
Previous ee-repo-ref: 7815dafb68d34ec5fbeb0645a095fbb6eae8d4a0
New ee-repo-ref: c9b043f2860fdae150c8c4bf03f3ec98b7f300e5
Automated by sync-ee-ref workflow.
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix: keep each dev server's session when worktrees share a host
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep dev auth cookies host-only on non-localhost hosts
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix: keep the shared dev auth cookie with ISOLATE_DEV_AUTH=0
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
