[ArtyomSavchenko]

No description provided.

[claude]

Review summary

This sync covers a lot of small fixes, and most of them come with focused unit tests, which is good to see.

Checked and look fine:

• Postgres \$in/\$nin with only null/undefined values, plus the inferType fix.
• The CockroachDB-safe primary key swap. The migration runner tracks migrations by name only, so editing the SQL of an already-applied migration doesn't re-run it.
• The withCurrentSpace refactor. It no longer mutates the model's actionProps, which is a real fix.
• The guard for attributes whose class has no domain.
• Excluding activity from printing.
• Removing the dead tracker Import action.
• Showing status colors per project type in the status filter.

Two inline comments:

1. Calendar (main concern): writing the Meet link into location means a user can't clear the location in Huly. The next incoming sync puts the link back. A test for that round trip is missing.
2. Contacts (minor): the e-mail regex accepts ?/&, so a stored channel value can add extra mailto: fields such as bcc or body.

No duplicated resource definitions or missing model migrations found.

[MichaelUray]

I reviewed the sync and the follow-up commit e72015e. The port itself looks clean: the 10 commits are the real upstream commits merged in d432c99, and for the 21 files changed on both sides the result matches an automatic 3-way merge (19 exactly; outcomingClient.ts only had an import-order conflict, resolved correctly). The Postgres $in/$nin null handling, the account migration, passing the current space from the command palette, the attribute-usage guard, the status filter colours, the print exclusion and the deadline/Import cleanups all look good, and their new tests call the real helpers. The mailto concern is fixed.

The calendar part of the follow-up still has problems, so I'd suggest fixing these before or right after merge:

1. Changed or removed conference links stay in Huly and get written back to Google (services/calendar/pod-calendar/src/conference.ts:49-53, sync.ts:490-492). Import an event with Meet link A and no location: Huly stores A as location. If the organizer replaces A with B or removes conferencing, the next incoming sync only reads Google's literal location, so Huly keeps A. The next unrelated edit in Huly then sees location out of sync and writes A into Google's real location field (calendar.ts:371 / outcomingClient.ts:289), so a removed meeting link comes back for all attendees. On the create path (e.g. after a 404 on update) the synthesized link is also written as a real Google location.
2. Most of upstream #11150 is effectively undone. The link is now only imported when the event is first created in Huly, so events synced before the deploy, and links added to existing Google events, never get it, and the fork diverges from upstream in these files (the next sync will conflict or bring the original behaviour back). One option: keep the conference link separately from the editable location, or import it on update only when the Huly location was never set (undefined, while a cleared location is ''), and send the same fix upstream.
3. The round-trip test does not run production code (services/calendar/pod-calendar/src/__tests__/conference.test.ts:92-101). It reimplements the incoming step as google.location ?? huly and never calls the outgoing updaters, so reverting the sync.ts fix keeps it green. Tests through the real incoming and outgoing paths for replacement, removal and a later unrelated edit would catch point 1.

Nits:

• The mailto branch cannot be reached today (Email/Telegram providers have no action), and the regex still accepts ,.
• tel: links now open in the same window; worth a quick check in the desktop app that this does not navigate the main window.
• Enabling tests in contact-resources also runs the fork-only person-id-filter.test.ts in CI for the first time.