fix(gmail): "hent nu" svarer selv, i stedet for at lade Cloudflare gøre det

En postkasse, der ikke svarer, holdt hele forespørgslen åben. Efter hundrede
sekunder gav Cloudflare op og sendte sin egen 502-side tilbage — HTML uden
fejlkode, så fladen kunne kun sige "Noget gik galt", og fejlloggen var tom,
fordi handleren aldrig nåede at vende tilbage.

Synk får nu sit eget budget på femogtyve sekunder. Løkken tjekker fristen mellem
hver besked, og hver opgave gemmes efterhånden som den laves, så en overskredet
frist er en kortere kørsel og ikke en tabt. Baggrundsjobbet beholder den fulde
tid; det har ingen, der venter.

Endepunkterne kan nu pege et andet sted hen, så testen kan stille en postkasse
op, der aldrig svarer, og se hvor lang tid der går. Budgettet kan sættes, så
den prøve koster et sekund frem for femogtyve.

Et budget på nul betyder standarden — ikke ingen tid. Uden det ville en
opsætning, der ikke satte feltet, lade fristen udløbe før første kald og gøre
hver eneste synk til et stilfærdigt nul.

Co-Authored-By: Claude Opus 5 <[email protected]>

internal/config/config.go

@@ -50,6 +50,16 @@ type Config struct {
 	// can read this environment can restart it, which is true of anybody with the
 	// host anyway. It is never sent to a browser: the settings page is told only
 	// whether it is set.
+	// GmailSyncBudget bounds one "fetch now", well under whatever proxy sits in
+	// front of this server. Configurable so a test can prove the budget is applied
+	// without spending the real one on every run.
+	GmailSyncBudget time.Duration
+
+	// GmailEndpoints redirects the Gmail calls at a server the tests control. Empty
+	// in production, where it means Google.
+	GmailTokenURL string
+	GmailAPIURL   string
+
 	PanelURL      string
 	PanelToken    string
 	PanelServerID string
@@ -85,6 +95,7 @@ func Load() (*Config, error) {
 		Dev:            envBool("VERDANDE_DEV", false),
 
 		GmailClientID:     env("VERDANDE_GMAIL_CLIENT_ID", ""),
+		GmailSyncBudget:   25 * time.Second,
 		PanelURL:          strings.TrimSuffix(env("VERDANDE_PANEL_URL", ""), "/"),
 		PanelToken:        env("VERDANDE_PANEL_TOKEN", ""),
 		PanelServerID:     env("VERDANDE_PANEL_SERVER_ID", ""),

internal/gmail/gmail.go

@@ -25,20 +25,50 @@ import (
 )
 
 const (
-	authEndpoint  = "https://accounts.google.com/o/oauth2/v2/auth"
-	tokenEndpoint = "https://oauth2.googleapis.com/token"
-	apiBase       = "https://gmail.googleapis.com/gmail/v1"
+	authEndpoint         = "https://accounts.google.com/o/oauth2/v2/auth"
+	defaultTokenEndpoint = "https://oauth2.googleapis.com/token"
+	defaultAPIBase       = "https://gmail.googleapis.com/gmail/v1"
 
 	// Read-only, and only mail. Not profile, not contacts, not send.
 	Scope = "https://www.googleapis.com/auth/gmail.readonly"
 )
 
+// Endpoints is where Google is.
+//
+// A field rather than a constant so a test can point it at a server it controls.
+// There is no other way to exercise what happens when Gmail is slow or refuses —
+// and "the sync comes back before whatever sits in front of this server gives up
+// on it" is exactly the kind of promise worth a test and impossible to make
+// against the real thing.
+//
+// The zero value is Google, so nothing outside a test has to know it exists.
+type Endpoints struct {
+	Token string
+	API   string
+}
+
+func (e Endpoints) token() string {
+	if e.Token != "" {
+		return e.Token
+	}
+	return defaultTokenEndpoint
+}
+
+func (e Endpoints) api() string {
+	if e.API != "" {
+		return e.API
+	}
+	return defaultAPIBase
+}
+
 // Config is the OAuth client the operator registered in Google Cloud. It belongs to
 // the instance rather than to a person: one registration, any number of users.
 type Config struct {
 	ClientID     string
 	ClientSecret string
 	RedirectURL  string
+	// Endpoints is Google unless a test says otherwise.
+	Endpoints Endpoints
 }
 
 func (c Config) Configured() bool {
@@ -128,7 +158,7 @@ func (c Config) Refresh(ctx context.Context, refreshToken string) (Token, error)
 }
 
 func (c Config) token(ctx context.Context, form url.Values) (Token, error) {
-	req, err := http.NewRequestWithContext(ctx, http.MethodPost, tokenEndpoint,
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.Endpoints.token(),
 		strings.NewReader(form.Encode()))
 	if err != nil {
 		return Token{}, err
@@ -171,6 +201,7 @@ func (c Config) token(ctx context.Context, form url.Values) (Token, error) {
 
 type Client struct {
 	accessToken string
+	endpoints   Endpoints
 	http        *http.Client
 }
 
@@ -178,6 +209,12 @@ func NewClient(accessToken string) *Client {
 	return &Client{accessToken: accessToken, http: &http.Client{Timeout: 30 * time.Second}}
 }
 
+// At points the client at another server. Tests only; the zero value is Google.
+func (c *Client) At(e Endpoints) *Client {
+	c.endpoints = e
+	return c
+}
+
 // Message is the part of a Gmail message a task is made from.
 type Message struct {
 	ID       string
@@ -196,7 +233,7 @@ func (c *Client) Profile(ctx context.Context) (string, error) {
 	var parsed struct {
 		EmailAddress string `json:"emailAddress"`
 	}
-	if err := c.get(ctx, apiBase+"/users/me/profile", &parsed); err != nil {
+	if err := c.get(ctx, c.endpoints.api()+"/users/me/profile", &parsed); err != nil {
 		return "", err
 	}
 	return parsed.EmailAddress, nil
@@ -220,7 +257,7 @@ func (c *Client) List(ctx context.Context, query string, max int) ([]string, err
 			ID string `json:"id"`
 		} `json:"messages"`
 	}
-	if err := c.get(ctx, apiBase+"/users/me/messages?"+v.Encode(), &parsed); err != nil {
+	if err := c.get(ctx, c.endpoints.api()+"/users/me/messages?"+v.Encode(), &parsed); err != nil {
 		return nil, err
 	}
 
@@ -251,7 +288,7 @@ func (c *Client) Get(ctx context.Context, id string) (Message, error) {
 			} `json:"headers"`
 		} `json:"payload"`
 	}
-	if err := c.get(ctx, apiBase+"/users/me/messages/"+url.PathEscape(id)+"?"+v.Encode(), &parsed); err != nil {
+	if err := c.get(ctx, c.endpoints.api()+"/users/me/messages/"+url.PathEscape(id)+"?"+v.Encode(), &parsed); err != nil {
 		return Message{}, err
 	}
 

internal/httpapi/gmail_handlers.go

@@ -2,6 +2,7 @@ package httpapi
 
 import (
 	"context"
+	"errors"
 	"net/http"
 	"strings"
 	"time"
@@ -10,6 +11,11 @@ import (
 	"github.com/kristianwind/verdande/internal/store"
 )
 
+// defaultGmailSyncBudget bounds one "fetch now" when nothing else says otherwise.
+// Well under the hundred seconds a proxy is willing to wait, so this server is the
+// one that answers a slow mailbox.
+const defaultGmailSyncBudget = 25 * time.Second
+
 // gmailConfig is the OAuth client this instance authorises with.
 //
 // The environment wins when it is set. A value in the environment is a deliberate
@@ -17,11 +23,20 @@ import (
 // applied on every recreate — and a form that could quietly override it would make
 // the manifest a lie. The stored pair is for the ordinary case, where nobody wants
 // to redeploy a container to paste a client id.
+// gmailClient is the API client, pointed wherever the config says Google is.
+func (s *Server) gmailClient(accessToken string) *gmail.Client {
+	return gmail.NewClient(accessToken).At(gmail.Endpoints{
+		Token: s.cfg.GmailTokenURL, API: s.cfg.GmailAPIURL,
+	})
+}
+
 func (s *Server) gmailConfig(ctx context.Context) gmail.Config {
+	endpoints := gmail.Endpoints{Token: s.cfg.GmailTokenURL, API: s.cfg.GmailAPIURL}
 	cfg := gmail.Config{
 		ClientID:     s.cfg.GmailClientID,
 		ClientSecret: s.cfg.GmailClientSecret,
 		RedirectURL:  s.cfg.GmailRedirectURL(),
+		Endpoints:    endpoints,
 	}
 	if cfg.ClientID != "" && cfg.ClientSecret != "" {
 		return cfg
@@ -153,7 +168,7 @@ func (s *Server) handleGmailCallback(w http.ResponseWriter, r *http.Request) {
 	// Best effort: a mailbox that connects but cannot name itself is still
 	// connected, and the status endpoint asks again when it finds no address —
 	// so this failing is a delay rather than a dead end.
-	email, err := gmail.NewClient(token.AccessToken).Profile(r.Context())
+	email, err := s.gmailClient(token.AccessToken).Profile(r.Context())
 	if err != nil {
 		s.log.Warn("gmail profile", "err", err)
 	}
@@ -185,8 +200,35 @@ func (s *Server) handleGmailCallback(w http.ResponseWriter, r *http.Request) {
 func (s *Server) handleGmailSyncNow(w http.ResponseWriter, r *http.Request) {
 	user := userFrom(r.Context())
 
-	created, err := s.SyncGmail(r.Context(), user)
+	// A budget of its own, well under what sits in front of this server.
+	//
+	// The sync fetches up to twenty-five messages one at a time, each with its own
+	// thirty-second timeout — so a slow mailbox can hold the request open for
+	// minutes. Cloudflare gives up at a hundred seconds and answers with its own
+	// 502 page, which is HTML: the browser gets no JSON, no error code and no
+	// message, and shows "something went wrong". Nothing reaches the error log
+	// either, because this handler never returned.
+	//
+	// Whatever was created by the deadline is kept: every task is committed as it
+	// is made, so a short budget is a shorter run rather than a lost one. The
+	// background job keeps the full one; it has nobody waiting.
+	// An unset budget is not a budget of nothing: a zero here would expire the
+	// context before the first call and make every sync a silent no-op.
+	budget := s.cfg.GmailSyncBudget
+	if budget <= 0 {
+		budget = defaultGmailSyncBudget
+	}
+	ctx, cancel := context.WithTimeout(r.Context(), budget)
+	defer cancel()
+
+	created, err := s.SyncGmail(ctx, user)
 	if err != nil {
+		if errors.Is(err, context.DeadlineExceeded) {
+			// Not a failure: the mailbox is slow or busy, some tasks were made, and
+			// the rest arrive on the next run ten minutes from now.
+			writeJSON(w, http.StatusOK, map[string]any{"created": created, "partial": true})
+			return
+		}
 		s.upstream(w, r, CodeGmailFailed, "sync gmail", err)
 		return
 	}
@@ -232,7 +274,7 @@ func (s *Server) SyncGmail(ctx context.Context, user *store.User) (int, error) {
 		}
 	}
 
-	client := gmail.NewClient(access)
+	client := s.gmailClient(access)
 	ids, err := client.List(ctx, query, 25)
 	if err != nil {
 		if err == gmail.ErrUnauthorized {
@@ -272,6 +314,11 @@ func (s *Server) SyncGmail(ctx context.Context, user *store.User) (int, error) {
 		if seen[id] {
 			continue
 		}
+		// A spent budget ends the run rather than turning into twenty-five failed
+		// fetches logged one at a time.
+		if ctx.Err() != nil {
+			return created, ctx.Err()
+		}
 		msg, err := client.Get(ctx, id)
 		if err != nil {
 			s.log.Warn("gmail get message", "err", err, "id", id)
@@ -454,7 +501,7 @@ func (s *Server) gmailProfile(ctx context.Context, user *store.User, settings ma
 		settings["expires_at"] = token.ExpiresAt.Unix()
 	}
 
-	email, err := gmail.NewClient(access).Profile(ctx)
+	email, err := s.gmailClient(access).Profile(ctx)
 	if err != nil {
 		return "", err
 	}

internal/httpapi/gmail_test.go

@@ -3,7 +3,9 @@ package httpapi
 import (
 	"net/http"
 	"net/http/httptest"
+	"net/url"
 	"testing"
+	"time"
 
 	"github.com/kristianwind/verdande/internal/config"
 )
@@ -26,6 +28,8 @@ func TestAGmailFailureSaysWhatGmailSaidAndIsRecorded(t *testing.T) {
 	ts := newTestServerWith(t, func(cfg *config.Config) {
 		cfg.GmailClientID = "test-client"
 		cfg.GmailClientSecret = "test-secret"
+		cfg.GmailTokenURL = google.URL + "/token"
+		cfg.GmailAPIURL = google.URL
 	})
 	ts.bootstrap(t)
 
@@ -68,3 +72,88 @@ func TestAGmailFailureSaysWhatGmailSaidAndIsRecorded(t *testing.T) {
 		t.Errorf("status = %v, want 502", row["status"])
 	}
 }
+
+// A slow mailbox must not hold the request open until something in front of this
+// server gives up on it.
+//
+// This is what was actually happening in production: each Gmail call has its own
+// thirty-second timeout, and the sync makes up to twenty-five of them one after the
+// other. Cloudflare stopped waiting at a hundred seconds and answered with its own
+// 502 page — HTML, so the browser got no code, no message and no error-log row, and
+// showed "something went wrong". The handler had never returned.
+func TestASlowMailboxDoesNotHangTheRequest(t *testing.T) {
+	// A Gmail that never answers.
+	blocked := make(chan struct{})
+	defer close(blocked)
+	google := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		select {
+		case <-blocked:
+		case <-r.Context().Done():
+		}
+	}))
+	defer google.Close()
+
+	ts := newTestServerWith(t, func(cfg *config.Config) {
+		cfg.GmailClientID = "test-client"
+		cfg.GmailClientSecret = "test-secret"
+		cfg.GmailTokenURL = google.URL + "/token"
+		cfg.GmailAPIURL = google.URL
+		// A second rather than the real twenty-five: what is being tested is that a
+		// budget is applied at all, and spending the production one on every run
+		// would put half a minute into the suite to learn the same thing.
+		cfg.GmailSyncBudget = time.Second
+	})
+	ts.bootstrap(t)
+
+	user, err := ts.db.UserByEmail(t.Context(), "[email protected]")
+	if err != nil {
+		t.Fatal(err)
+	}
+	// A token that has not expired, so the run goes straight to the message list
+	// rather than stopping at a refresh.
+	if err := ts.db.SetUserSettings(t.Context(), user.ID, "gmail", map[string]any{
+		"refresh_token": "r", "access_token": "a", "trigger": "starred",
+		"expires_at": time.Now().Add(time.Hour).Unix(),
+	}); err != nil {
+		t.Fatal(err)
+	}
+
+	// Its own client, because the shared one gives up after ten seconds — which is
+	// shorter than the budget being tested and would prove nothing either way.
+	req, _ := http.NewRequest("POST", ts.URL+"/api/v1/gmail/sync", nil)
+	req.Header.Set("Sec-Fetch-Site", "same-origin")
+	for _, c := range ts.client.Jar.Cookies(mustParse(t, ts.URL)) {
+		req.AddCookie(c)
+	}
+
+	started := time.Now()
+	resp, err := (&http.Client{Timeout: 90 * time.Second}).Do(req)
+	if err != nil {
+		t.Fatalf("the request never came back (%v after %s); the sync has no budget of "+
+			"its own, and whatever sits in front of this server answers for it",
+			err, time.Since(started).Round(time.Second))
+	}
+	defer resp.Body.Close()
+	took := time.Since(started)
+
+	// Cloudflare stops waiting at a hundred seconds. Anything close to that is the
+	// bug, whatever status eventually arrives.
+	// Comfortably over the one-second budget and far under any proxy's patience.
+	if took > 10*time.Second {
+		t.Errorf("the sync took %s; the budget is not being applied",
+			took.Round(time.Second))
+	}
+	if resp.StatusCode >= 500 && resp.StatusCode != http.StatusBadGateway {
+		t.Errorf("status %d — a slow mailbox is not this server breaking", resp.StatusCode)
+	}
+	t.Logf("came back in %s with %d", took.Round(time.Millisecond), resp.StatusCode)
+}
+
+func mustParse(t *testing.T, raw string) *url.URL {
+	t.Helper()
+	u, err := url.Parse(raw)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return u
+}