Add a fake Basecamp for testing basecamp connect end to end - #822

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/faults_test.go, line 151:

<comment>This substring check passes even if `performed_by_id` appears at the wrong JSON level, so it does not validate the event shape. Decode the event JSON and assert the field's presence and value.

(Based on your team's feedback about structured JSON assertions.) .</comment>

<file context>
@@ -0,0 +1,181 @@
+	page, body := poll("since=0&exclude_performers=self")
+	require.Len(t, page.Events, 1)
+	assert.Equal(t, first.ID, page.Events[0].ID)
+	assert.Contains(t, body, `"performed_by_id":null`)
+	assert.Equal(t, s.url()+"/999/events.json?exclude_performers=self&position=fake-1", page.Next)
+
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/faults_test.go, line 94:

<comment>The worker calls `do`, whose `require.NoError` exits only that goroutine on request errors, so the test blocks forever waiting on `answered`. Return the request error through the channel and assert from the test goroutine.</comment>

<file context>
@@ -0,0 +1,181 @@
+	gate := s.Gate(fakebasecamp.RouteProjects)
+	answered := make(chan int, 1)
+	go func() {
+		status, _, _ := do(t, s, http.MethodGet, "/999/projects.json", bearer, "")
+		answered <- status
+	}()
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/reads_test.go, line 36:

<comment>These JSON assertions search raw text, so they can pass when `boss` or `identity.id` is missing from its expected location. Decode each response and assert the structured fields instead.

(Based on your team's feedback about structured JSON assertions.) </comment>

<file context>
@@ -0,0 +1,160 @@
+	assert.True(t, ok)
+	assert.Equal(t, fakebasecamp.AgentID, id, "the sgid names the person, as a mention's does")
+	_, _, body := do(t, s, http.MethodGet, "/999/my/profile.json", bearer, "")
+	assert.Contains(t, body, `"boss":{"id":26909558,"name":"Operator"}`)
+
+	client, err := agent.People().Get(bounded(t), 1003)
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/feed.go, line 331:

<comment>A valid cursor of `9223372036854775807` overflows `cursor+1`, so the search restarts at the first event and replays older events. Search for the first event whose ID is greater than `cursor` without incrementing it.</comment>

<file context>
@@ -0,0 +1,350 @@
+		cursor = n
+	}
+
+	start, _ := slices.BinarySearchFunc(poll, cursor+1, func(ev Event, id int64) int { return cmp.Compare(ev.ID, id) })
+	end := min(start+c.s.opts.pageSize, len(poll))
+	rows := []pollRow{}
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/feed.go, line 332:

<comment>`WithPageSize(0)` leaves `end == start`, so every `next` URL repeats the same cursor forever; a negative value instead panics when slicing `poll[start:end]`. Clamp or reject non-positive sizes before computing the end.</comment>

<file context>
@@ -0,0 +1,350 @@
+	}
+
+	start, _ := slices.BinarySearchFunc(poll, cursor+1, func(ev Event, id int64) int { return cmp.Compare(ev.ID, id) })
+	end := min(start+c.s.opts.pageSize, len(poll))
+	rows := []pollRow{}
+	at := cursor
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/server.go, line 173:

<comment>A concurrent second `Close` returns as soon as it sees `closed`, before the first call finishes `srv.Close()` and `cables.Wait()`. Make later callers wait for shutdown completion.</comment>

<file context>
@@ -0,0 +1,651 @@
+// it does. It is safe to call more than once.
+func (s *Server) Close() {
+	s.mu.Lock()
+	if s.closed {
+		s.mu.Unlock()
+		return
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/server.go, line 97:

<comment>`WithPingInterval(0)` and negative values reach `time.NewTicker` in the cable goroutine and panic when a cable opens. Reject nonpositive intervals while applying options.</comment>

<file context>
@@ -0,0 +1,651 @@
+// WithPingInterval sets how often the cable pings. The default is Action
+// Cable's three seconds, which is what eventfeed's staleness rule is built
+// around.
+func WithPingInterval(d time.Duration) Option { return func(o *options) { o.pingInterval = d } }
+
+// WithListener serves on l instead of a fresh loopback port.
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At e2e/fakebasecamp/main.go, line 137:

<comment>Quote the formatted message before logging so control characters in a request path cannot forge extra log entries.</comment>

<file context>
@@ -0,0 +1,151 @@
+
+func (*logReporter) Helper() {}
+
+func (*logReporter) Errorf(format string, args ...any) { log.Printf(format, args...) }
+
+func (r *logReporter) Cleanup(fn func()) {
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/doc.go, line 23:

<comment>Nonmembers do not get 404s from the feed lanes: the poll endpoint omits inaccessible events, and the cable skips them. Distinguish those behaviors from direct reads so tests do not rely on the wrong contract.</comment>

<file context>
@@ -0,0 +1,56 @@
+//
+// Requests are authorized as Basecamp authorizes them: a bearer token names a
+// person, and a person reads what their projects hold. A project the caller
+// is not a member of, and anything in it, answers 404, on the reads and on
+// both lanes of the feed.
+//
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At internal/connector/fakebasecamp/recordings.go, line 20:

<comment>`Assignments.AddedPersonIDs` is capped at `MaxAssignmentPages` and stops as soon as it finds the event, so this says an empty page is the only stopping condition. Describe the bounded search instead.</comment>

<file context>
@@ -15,7 +15,11 @@ import (
-	// recordingEventsPage is how many history entries a page carries.
+	// recordingEventsPage is how many history entries a page carries:
+	// fifteen, the first page Basecamp's geared pagination serves. The
+	// connector walks pages until one comes back empty and never counts on
+	// the size, so the fake keeps one size for every page; a small one
+	// makes a walk past the first page cheap to set up.
</file context>