Skip to content

fix(scans): queue accepted hackers who can't check in as walk-ins - #360

Merged
balebbae merged 2 commits into
mainfrom
fix/walk-in-unclaimed-spots
Nov 7, 2026
Merged

balebbae merged 2 commits into
mainfrom
fix/walk-in-unclaimed-spots

Conversation

@balebbae

@balebbae balebbae commented Nov 7, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

An accepted hacker whose RSVP isn't confirmed (pending, declined, withdrawn or forfeited) was refused by both scanners:

  • Check-in refused them for the RSVP.
  • Walk-in refused them for being accepted ("already accepted: scan them at check-in").

So volunteers would send them back and forth, and only a super admin's Admit by email could let them in. In prod tonight that's 459 people: 395 pending, 51 declined, 13 withdrawn.

This change makes the walk-in scan refuse only the people check-in would let in, using one shared function (checkInRefusal), and queues everyone else. The two scanners can no longer disagree.

Person Check-in Walk-in before Walk-in after
Accepted, RSVP confirmed ✅ in 409 409 "has a spot: scan them at check-in"
Accepted, RSVP pending / declined / withdrawn / forfeited 403 409 (stuck) ✅ queued
Rejected / draft / never applied 403 queued queued (unchanged)
Promoted walk-in ✅ in 409 409 "has a spot"

Changes

  • cmd/api/scans.go
    • The check-in rules move into checkInRefusal. The check-in branch and the walk-in branch both call it.
    • The walk-in branch now checks the queue first (the cheap exit on a re-scan), then refuses only when checkInRefusal says they can check in.
    • Promoted walk-ins are now exempt from every RSVP check. Before, a hacker who withdrew could join the queue and be promoted, but would still be refused at check-in.
  • internal/store/walkins.go: removes Enqueue's "already accepted → don't queue" shortcut. The handler now owns the rule. Without this, the handler change alone would record a walk-in scan and answer 201 without queueing anyone.
  • claude.md: the walk-in rule is updated.

As before, queueing makes the application waitlisted. Promotion makes it accepted again. The portal already shows a queued walk-in their position, and they get the queued email.

Also in this PR: a chat UI tweak

  • Light-mode chat card: the background is less see-through (bg-surface/40 → /70), so the starter questions are readable over the page. Dark mode keeps its own 40% from index.css.
  • Extra starter questions: on hover they now fill with solid blue and white text, like the schedule starter, instead of a faint tint.

Tests

  • Handler tests:
    • Accepted hackers with a pending, declined, withdrawn or forfeited RSVP are queued and get a queue position.
    • Walk-in refuses with 409 for a confirmed RSVP, a promoted walk-in, and a pending RSVP when the requirement is off.
    • A promoted walk-in with a declined, withdrawn or forfeited RSVP checks in.
  • Store integration test (TestIntegrationWalkInsEnqueueUnclaimedAcceptance): an accepted hacker with a pending RSVP is queued, is waitlisted while queued, and is then promoted with PromotedWalkIn set. It runs in the db-integration CI job; it wasn't run locally.
  • go test -race ./... passes locally.

Deploying (event starts 6am Nov 7)

This branch is based on v0.27.0, not main. Do not merge release PR #358 this weekend: it ships #357 and #359, and #359 needs migration 124, which isn't applied to prod. The container never runs migrations.

  1. Once CI is green, tag this branch's head (18173befa) as v0.27.1 and push the tag. harp-prod-us-central1 builds and swaps only the image, so min 3 / max 40 / CPU always on stay as they are. No migration is needed.
  2. Merge this PR into main so the fix isn't lost. release-please's manifest still says 0.27.0, so check the version it proposes next.

Things to know

  • The admin "accepted" count goes down by however many of these people join the queue. That's accurate: they're waitlisted until promoted.
  • Don't run a decision release during the event. A queued person's status (waitlisted) differs from their released_status (accepted), so a release would send them a decision email. Existing walk-ins already behave this way.
  • Some emails will still treat them as accepted. Email audiences filter on the decision the hacker last saw, so a queued person still matches "accepted" audiences.

🤖 Generated with Claude Code

An accepted hacker whose RSVP is pending, declined, withdrawn or forfeited
was refused by both scanners: check-in for the RSVP, walk-in for being
accepted. The walk-in scan now refuses only people check-in would let in,
using one shared rule, and queues everyone else. Promotion from the queue
now also clears a withdrawn or forfeited RSVP at check-in.
@balebbae
balebbae merged commit 36393b8 into main Nov 7, 2026
8 checks passed
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