beads/cmd/bd/update_proxied_server_test.go
CI Bot 67d38d2620 Trim this series' comments toward the repo's own density
Ten agents over 216 files, one directory-balanced group each, working to a
rubric that named what to cut (duplication narration, restatement of the next
line, the governing rule echoed in every command file, throat-clearing) and
what to keep (traps, invariants, security, tracker ids, behaviour changes, and
the "why a naive simplification destroys this assertion" rationale).

Net 1,523 lines. Density against this repo's own per-directory norm on main:

  cmd/bd              33.6% -> 27.3%   (main 10.7%)
  internal/storage    26.1% -> 23.1%   (main 14.1%)
  backend/conformance 34.0% -> 29.6%   (main 22.8%)
  internal/httpapi    30.9% -> 27.8%   (main 30.6%, now below)
  overall             34.7% -> 31.6%

Every agent landed at 18-25% rather than the 30% asked for, and every one gave
the same reason independently: what remained was on the keep list. Ten separate
judgements agreeing is the most useful result here — the prose is dense, but it
is mostly load-bearing, and the shortfall is the honest answer rather than a
failure to try.

TWO REVERTS, both to protect something worth more than the lines.

issueops/ is back to its pre-trim state, giving up 610 lines. The conformance
contracts cite the leaf docs BY LINE — "pins treewalker.go:52-66" — 155 times.
Trimming the cited files silently repointed those at whatever text moved into
those numbers, across agent boundaries so no agent could see it. A semantic
remapper recovered 141 of them, but 23 needed judgement and several landed on
struct fields. The leaf docs are the spec the contracts are written against;
they are the one place in this change where dense prose IS the deliverable, and
keeping them stable is worth more than the 610 lines. tree_walker_contract.go
and version_reconciler_contract.go went back for the same reason: an agent had
deleted a paragraph containing a citation, so positional restoration could not
align them.

After the reverts, exactly 5 citations point into a still-trimmed file and 2 of
those are weak — both in cmd/bd/config_guard_test.go, which this series never
touched and whose citations were already wrong on main (init.go:361 pointed at
a closing brace; the guard it names is at init.go:1499).

VERIFICATION. A guard strips every comment and all whitespace from each of the
193 modified files at HEAD and in the worktree and requires the two to be
byte-identical, so a code change cannot hide in a comment diff. It handles `//`
inside string and raw-string literals, and was tested against a deliberate code
change before being trusted. gofmt clean, go build, go vet, golangci-lint 0
issues, and the two AST-based tests that PARSE trimmed files — the timed
provider binding pin over claim.go and the serve database-source gate over
serve.go — both pass.

One real defect found while trimming: checkDatabaseSource's doc block ran into
`// sourceRoles is ...` with no blank line, so godoc attached thirty-one lines
of the wrong prose to func sourceRoles and left checkDatabaseSource
undocumented. Fixed by separating them.
2026-08-06 08:06:59 +00:00

310 lines
12 KiB
Go

package main
import (
"context"
"encoding/json"
"errors"
"fmt"
"io"
"os"
"strings"
"testing"
mysql "github.com/go-sql-driver/mysql"
"github.com/steveyegge/beads/internal/storage"
"github.com/steveyegge/beads/internal/types"
"github.com/steveyegge/beads/issueops"
)
// The plain proxied update joined the claim on issueops.Lifecycle, so these pin
// the same three things its sibling in update_claim_proxied_test.go pins: the
// request this surface makes, the roles it reaches them through, and the fact
// that it opens no unit of work of its own.
func captureStderrDuring(t *testing.T, fn func()) string {
t.Helper()
old := os.Stderr
r, w, err := os.Pipe()
if err != nil {
t.Fatalf("os.Pipe: %v", err)
}
os.Stderr = w
done := make(chan string, 1)
go func() {
data, _ := io.ReadAll(r)
done <- string(data)
}()
defer func() {
os.Stderr = old
_ = r.Close()
}()
fn()
_ = w.Close()
return <-done
}
func serializationFailure() error {
return &mysql.MySQLError{Number: 1213, Message: "serialization failure, transaction rolled back"}
}
func TestProxiedUpdateRunsOnTheLifecycleContract(t *testing.T) {
before := &types.Issue{ID: "bd-1", Status: types.StatusOpen, Title: "Original"}
after := &types.Issue{ID: "bd-1", Status: types.StatusOpen, Title: "Renamed", Labels: []string{"keep"}}
p := claimRoleFixture(t, before, issueops.UpdateResult{Issue: after, Changed: true}, nil)
oldActor := actor
actor = "agent"
t.Cleanup(func() { actor = oldActor })
in := &updateInput{fields: map[string]any{"title": "Renamed"}}
got, fail, err := applyUpdateProxiedOne(context.Background(), "bd-1", in)
if err != nil || fail != nil {
t.Fatalf("applyUpdateProxiedOne: err = %v, fail = %+v", err, fail)
}
if got == nil || got.Title != "Renamed" {
t.Fatalf("updated issue = %+v, want the renamed row", got)
}
// Only the claim strips labels, to keep its published response shape; a
// plain update carries them, as the direct route's does.
if len(got.Labels) != 1 || got.Labels[0] != "keep" {
t.Errorf("labels = %v, wanted the contract's hydrated labels kept on a plain update", got.Labels)
}
req := p.lifecycle.request
if req.Claim {
t.Error("a plain update reached the contract as a claim")
}
if req.Actor != "agent" || req.IssueID != "bd-1" || !req.Patch.Title.Set || req.Patch.Title.Value != "Renamed" {
t.Errorf("request = %+v, want a title edit of bd-1 by agent", req)
}
// The commit message this path has always written, carried on the request
// instead of being spelled at the commit site.
if req.Provenance != "bd: update bd-1" {
t.Errorf("Provenance = %q, want the message the proxied update has always written", req.Provenance)
}
// A CLI update resolves either plane; only the HTTP surface restricts it.
if req.IssuePlaneOnly {
t.Error("the CLI update restricted the plane; `bd update` has always resolved a wisp id")
}
if p.uows != 0 {
t.Errorf("the update path opened %d units of work; the contract owns the transaction", p.uows)
}
}
// TestProxiedUpdatePatchRoutesMergeOpsAsOperations locks in the lost-update fix
// on the proxied-server route: metadata edits and note appends must reach the
// contract as merge OPERATIONS, never as values pre-merged from the pre-read
// row (a read from an earlier transaction — merging there erased keys a
// concurrent writer committed after it).
func TestProxiedUpdatePatchRoutesMergeOpsAsOperations(t *testing.T) {
current := &types.Issue{
ID: "bd-spec-1",
Status: types.StatusOpen,
Notes: "existing notes",
Metadata: json.RawMessage(`{"existing":"yes"}`),
}
t.Run("append_notes", func(t *testing.T) {
patch := mustProxiedUpdatePatch(t, &updateInput{
fields: map[string]any{},
hasAppendNotes: true,
appendNotes: "appended",
}, current)
if !patch.AppendNotes.Set || patch.AppendNotes.Value != "appended" {
t.Errorf("AppendNotes = %+v, want the raw append text", patch.AppendNotes)
}
if patch.Notes.Set {
t.Errorf("Notes = %+v: notes must NOT be pre-merged from the pre-read row", patch.Notes)
}
})
t.Run("merge_metadata", func(t *testing.T) {
patch := mustProxiedUpdatePatch(t, &updateInput{
fields: map[string]any{},
mergeMetadataIn: json.RawMessage(`{"new":"key"}`),
}, current)
if !patch.Metadata.Merge.Set || string(patch.Metadata.Merge.Value) != `{"new":"key"}` {
t.Errorf("Metadata.Merge = %+v, want the raw incoming JSON", patch.Metadata.Merge)
}
if patch.Metadata.Replace.Set {
t.Errorf("Metadata.Replace = %+v: metadata must NOT be pre-merged from the pre-read row", patch.Metadata.Replace)
}
})
t.Run("set_and_unset_metadata", func(t *testing.T) {
patch := mustProxiedUpdatePatch(t, &updateInput{
fields: map[string]any{},
setMetadata: []string{"tier=gold"},
unsetMetadata: []string{"existing"},
}, current)
if got := string(patch.Metadata.Set["tier"]); got != `"gold"` {
t.Errorf("Metadata.Set[tier] = %s, want the JSON string value", got)
}
if len(patch.Metadata.Unset) != 1 || patch.Metadata.Unset[0] != "existing" {
t.Errorf("Metadata.Unset = %v, want [existing]", patch.Metadata.Unset)
}
if patch.Metadata.Replace.Set {
t.Errorf("Metadata.Replace = %+v: metadata must NOT be pre-merged from the pre-read row", patch.Metadata.Replace)
}
})
t.Run("clear_defer_status_resolved_from_the_pre_read_row", func(t *testing.T) {
deferred := &types.Issue{ID: "bd-spec-2", Status: types.StatusDeferred}
patch := mustProxiedUpdatePatch(t, &updateInput{fields: map[string]any{}, clearDeferStatus: true}, deferred)
if !patch.Status.Set || patch.Status.Value != types.StatusOpen {
t.Errorf("Status = %+v, want open (clearDeferStatus on a deferred issue)", patch.Status)
}
open := &types.Issue{ID: "bd-spec-3", Status: types.StatusBlocked}
if patch := mustProxiedUpdatePatch(t, &updateInput{fields: map[string]any{}, clearDeferStatus: true}, open); patch.Status.Set {
t.Errorf("Status = %+v, want unset: --defer=\"\" must not clobber a non-deferred status", patch.Status)
}
})
}
func mustProxiedUpdatePatch(t *testing.T, in *updateInput, before *types.Issue) issueops.IssuePatch {
t.Helper()
patch, err := proxiedUpdatePatch(in, before)
if err != nil {
t.Fatalf("proxiedUpdatePatch: %v", err)
}
return patch
}
// TestProxiedUpdateCarriesForce pins the proxied path's translation of
// `--force`. An earlier attempt was reverted for exactly this missing mapping:
// the proxied caller built a request that never carried the override, so a
// shared policy check refused the close with no way for the user to say
// otherwise. The assignee half only applies to an assignee edit, so
// `--force -s closed` asks for the close-policy half alone.
func TestProxiedUpdateCarriesForce(t *testing.T) {
closing := map[string]any{"status": string(types.StatusClosed)}
forced := recordProxiedUpdateRequest(t, &updateInput{fields: closing, force: true})
if !forced.ForceClosePolicy {
t.Error("ForceClosePolicy = false with --force")
}
if forced.ForceAssigneeTransfer {
t.Error("ForceAssigneeTransfer = true without an assignee edit to authorize")
}
unforced := recordProxiedUpdateRequest(t, &updateInput{fields: closing})
if unforced.ForceClosePolicy {
t.Error("ForceClosePolicy = true without --force")
}
transfer := recordProxiedUpdateRequest(t, &updateInput{
fields: map[string]any{"assignee": "thief"}, force: true,
})
if !transfer.ForceAssigneeTransfer || !transfer.ForceClosePolicy {
t.Errorf("request = %+v, want --force to carry both halves on an assignee edit", transfer)
}
}
// TestProxiedUpdateCarriesConditionalGuards pins the bd-wsqvw guards onto the
// request. The contract runs the compare-and-set inside the mutation
// transaction, the only place it can be atomic with the write it gates.
func TestProxiedUpdateCarriesConditionalGuards(t *testing.T) {
holder, status := "alice", string(types.StatusInProgress)
req := recordProxiedUpdateRequest(t, &updateInput{
fields: map[string]any{"priority": 0},
ifAssignee: &holder,
ifStatus: &status,
})
if req.ExpectedAssignee == nil || *req.ExpectedAssignee != "alice" {
t.Errorf("ExpectedAssignee = %v, want alice", req.ExpectedAssignee)
}
if req.ExpectedStatus == nil || *req.ExpectedStatus != types.StatusInProgress {
t.Errorf("ExpectedStatus = %v, want in_progress", req.ExpectedStatus)
}
}
func recordProxiedUpdateRequest(t *testing.T, in *updateInput) issueops.UpdateRequest {
t.Helper()
before := &types.Issue{ID: "bd-1", Status: types.StatusOpen}
p := claimRoleFixture(t, before, issueops.UpdateResult{Issue: before, Changed: true}, nil)
if _, fail, err := applyUpdateProxiedOne(context.Background(), "bd-1", in); err != nil || fail != nil {
t.Fatalf("applyUpdateProxiedOne: err = %v, fail = %+v", err, fail)
}
return p.lifecycle.request
}
// TestProxiedUpdateExhaustedConflictsFailLoudly proves a write that never lands
// cannot exit as a success: when the contract spends its retry budget losing
// Dolt's commit-time merge, the command reports the failure instead of printing
// "✓ Updated" (a recorded failure suppresses the success line and drives the
// non-zero exit via reportUpdateFailures in runUpdateProxiedServer).
func TestProxiedUpdateExhaustedConflictsFailLoudly(t *testing.T) {
before := &types.Issue{ID: "bd-retry-2", Status: types.StatusOpen}
claimRoleFixture(t, before, issueops.UpdateResult{}, serializationFailure())
var (
got *types.Issue
fail *updateIDFailure
err error
)
stderr := captureStderrDuring(t, func() {
in := &updateInput{fields: map[string]any{"title": "never lands"}}
got, fail, err = applyUpdateProxiedOne(context.Background(), "bd-retry-2", in)
})
if err != nil {
t.Fatalf("applyUpdateProxiedOne returned a hard error: %v", err)
}
if fail == nil || got != nil {
t.Fatalf("fail=%v issue=%v: a write that never landed must be reported as a failure", fail, got)
}
if fail.GuardMismatch {
t.Error("exhausted conflicts must not masquerade as a guard mismatch (that would exit 13 and tell scripts not to retry)")
}
if !strings.Contains(stderr, "retries exhausted") {
t.Errorf("stderr = %q, want a loud retries-exhausted failure", stderr)
}
}
// TestProxiedUpdateGuardMismatchExitsThirteen keeps the one failure class that
// gets its own exit code: a stale --if-assignee/--if-status wrote nothing and
// retrying is pointless, so the batch must exit 13 rather than 1.
func TestProxiedUpdateGuardMismatchExitsThirteen(t *testing.T) {
before := &types.Issue{ID: "bd-guard-1", Status: types.StatusOpen}
claimRoleFixture(t, before, issueops.UpdateResult{},
fmt.Errorf("%w: bd-guard-1 has status %q, expected %q", storage.ErrStatusMismatch, "open", "in_progress"))
stale := string(types.StatusInProgress)
var fail *updateIDFailure
stderr := captureStderrDuring(t, func() {
_, fail, _ = applyUpdateProxiedOne(context.Background(), "bd-guard-1", &updateInput{
fields: map[string]any{"priority": 0}, ifStatus: &stale,
})
})
if fail == nil || !fail.GuardMismatch {
t.Fatalf("fail = %+v, want a guard-mismatch verdict", fail)
}
if !strings.Contains(fail.Error, "precondition failed") {
t.Errorf("fail.Error = %q, want it named as a precondition failure", fail.Error)
}
if !strings.Contains(stderr, "status mismatch") {
t.Errorf("stderr = %q, want the machine-greppable sentinel text", stderr)
}
}
// TestProxiedUpdateAbortsTheBatchOnCancellation is the plain-update half of the
// claim's own cancellation rule: SIGINT cancels bd's root context, every
// remaining id would fail the same way, so the loop aborts instead of recording
// one "context canceled" failure per id.
func TestProxiedUpdateAbortsTheBatchOnCancellation(t *testing.T) {
for _, cancellation := range []error{context.Canceled, context.DeadlineExceeded} {
t.Run(cancellation.Error(), func(t *testing.T) {
before := &types.Issue{ID: "bd-1", Status: types.StatusOpen}
claimRoleFixture(t, before, issueops.UpdateResult{}, fmt.Errorf("update bd-1: %w", cancellation))
got, fail, err := applyUpdateProxiedOne(context.Background(), "bd-1",
&updateInput{fields: map[string]any{"title": "interrupted"}})
if !errors.Is(err, cancellation) {
t.Fatalf("err = %v, want %v returned so the batch aborts", err, cancellation)
}
if got != nil || fail != nil {
t.Errorf("issue = %+v, fail = %+v: cancellation is not a per-id verdict", got, fail)
}
})
}
}