fix(validators): distinguish a missing package from a missing version on PyPI/NPM (#1411)

Fixes #553.

The PyPI and npm validators fetch the version-specific metadata endpoint
(`/pypi/{name}/{version}/json` and the npm equivalent) and report any
non-200 as `<pkg> not found`. That endpoint 404s for two different
reasons though: the package doesn't exist, or it exists and only the
*version* is missing (e.g. a release that hasn't propagated yet). So a
valid package reads as gone:

```diff
- PyPI package 'requests' not found (status: 404)
+ PyPI package 'requests' exists, but version '99.99.99' was not found (status: 404). A newly published release can take a moment to appear on PyPI. Wait and retry, or publish version '99.99.99' before registering it
```

On a version 404 it probes the package-level endpoint
(`/pypi/{name}/json`, `/{name}`) with a `HEAD` to tell the two apart,
and reports 429/5xx as transient rather than "not found". `HEAD` so it
reads a status without pulling the whole packument. The probe carries
its own 3s deadline: it only refines the error message, so a hung probe
must not stretch the validator past the ~10s-per-registry budget the
publish path assumes. Same probe-and-classify shape as the existing
cargo validator.

The fetch is split into `validatePyPIPackage` / `validateNPMPackage`
behind `export_test.go` so the branches are testable with `httptest`,
the same seam cargo uses.

### Testing
- [x] hermetic `httptest` for the status matrix (version-missing,
package-missing, 5xx, 429, inconclusive probe, positive path, scoped
npm); the mocks pin the expected method per endpoint (`GET` fetch,
`HEAD` probe)
- [x] hermetic deadline test: a hung probe is cut off at ~3s and
reported as inconclusive, not "not found"
- [x] existing live package tests still pass, now also hitting the HEAD
probe; on a live 429/5xx they now `t.Skip` as inconclusive instead of
flaking (deliberate: CI runs these against the real registries with no
short-mode gating)
- [x] `gofmt` / `go vet` / `golangci-lint` clean

Out of scope: auto-retry for the propagation race, and SSRF
redirect-pinning parity for the pypi/npm clients (pre-existing, shared
with nuget).

internal/validators/registries/npm_test.go

@@ -2,13 +2,186 @@ package registries_test
 
 import (
 	"context"
+	"fmt"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
 	"testing"
+	"time"
 
 	"github.com/modelcontextprotocol/registry/internal/validators/registries"
 	"github.com/modelcontextprotocol/registry/pkg/model"
 	"github.com/stretchr/testify/assert"
 )
 
+// newNPMMock stands in for registry.npmjs.org: it routes the version fetch and
+// package probe by path shape and returns the given statuses (versionBody used on 200).
+func newNPMMock(versionStatus int, versionBody string, packageStatus int) *httptest.Server {
+	return httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		// Route on the escaped path: a scoped name like @scope/pkg is sent as the
+		// single segment @scope%2Fpkg, and decoding it would wrongly split it in two.
+		parts := strings.Split(strings.Trim(r.URL.EscapedPath(), "/"), "/")
+		// Pin the method per endpoint (GET fetch, HEAD probe) so a method
+		// regression in the validator surfaces as a 405 instead of passing.
+		switch len(parts) {
+		case 2: // /{name}/{version}
+			if r.Method != http.MethodGet {
+				w.WriteHeader(http.StatusMethodNotAllowed)
+				return
+			}
+			if versionStatus != http.StatusOK {
+				w.WriteHeader(versionStatus)
+				return
+			}
+			w.Header().Set("Content-Type", "application/json")
+			_, _ = io.WriteString(w, versionBody)
+		case 1: // /{name}  (package-existence probe)
+			if r.Method != http.MethodHead {
+				w.WriteHeader(http.StatusMethodNotAllowed)
+				return
+			}
+			w.WriteHeader(packageStatus)
+		default:
+			w.WriteHeader(http.StatusInternalServerError)
+		}
+	}))
+}
+
+// TestValidateNPM_VersionNotYetVisible is the #553 regression for npm: version 404s
+// while the package exists, so the error must report a missing version, not a missing package.
+func TestValidateNPM_VersionNotYetVisible(t *testing.T) {
+	ctx := context.Background()
+	mock := newNPMMock(http.StatusNotFound, "", http.StatusOK)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "9.9.9"}
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "exists, but version '9.9.9'", "package-exists/version-missing must be distinguished from package-missing")
+}
+
+// TestValidateNPM_PackageMissing: both endpoints 404, so "not found" is correct.
+func TestValidateNPM_PackageMissing(t *testing.T) {
+	ctx := context.Background()
+	mock := newNPMMock(http.StatusNotFound, "", http.StatusNotFound)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "1.0.0"}
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "not found")
+	assert.NotContains(t, err.Error(), "exists, but version", "a genuinely missing package must not claim the version exists")
+}
+
+// TestValidateNPM_TransientUpstream: a 5xx on the version fetch must be retryable,
+// not "not found".
+func TestValidateNPM_TransientUpstream(t *testing.T) {
+	ctx := context.Background()
+	mock := newNPMMock(http.StatusBadGateway, "", http.StatusOK)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "1.0.0"}
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "transient")
+	assert.NotContains(t, err.Error(), "not found", "transient upstream errors must not be reported as 'not found'")
+}
+
+// TestValidateNPM_VersionNotFoundProbeInconclusive: version 404 plus a transient
+// probe (503) leaves existence undetermined, so the validator must not say "not found".
+func TestValidateNPM_VersionNotFoundProbeInconclusive(t *testing.T) {
+	ctx := context.Background()
+	mock := newNPMMock(http.StatusNotFound, "", http.StatusServiceUnavailable)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "1.0.0"}
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "transient")
+	assert.NotContains(t, err.Error(), "not found", "an inconclusive probe must not assert the package is missing")
+}
+
+// TestValidateNPM_ProbeDeadlineBounded: a hung probe must be cut off by the
+// probe's own short deadline instead of riding out the client's full 10s
+// timeout, and the cutoff must read as inconclusive rather than "not found".
+func TestValidateNPM_ProbeDeadlineBounded(t *testing.T) {
+	ctx := context.Background()
+	mock := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		parts := strings.Split(strings.Trim(r.URL.EscapedPath(), "/"), "/")
+		if len(parts) == 2 { // version fetch
+			w.WriteHeader(http.StatusNotFound)
+			return
+		}
+		// Package probe: hang until the client gives up.
+		<-r.Context().Done()
+	}))
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "1.0.0"}
+	start := time.Now()
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	elapsed := time.Since(start)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "transient", "a probe cut off by its deadline is inconclusive, not 'not found'")
+	assert.Less(t, elapsed, 8*time.Second, "a hung probe must be bounded by the probe deadline, not the client timeout")
+}
+
+// TestValidateNPM_ScopedVersionNotYetVisible covers scoped names (@scope/name), the
+// escaped-single-segment case, through the version fetch and the HEAD probe.
+func TestValidateNPM_ScopedVersionNotYetVisible(t *testing.T) {
+	ctx := context.Background()
+	mock := newNPMMock(http.StatusNotFound, "", http.StatusOK)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "@scope/demo", Version: "9.9.9"}
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "exists, but version '9.9.9'", "scoped names must route correctly through both endpoints")
+}
+
+// TestValidateNPM_VersionEndpointRateLimited: a 429 on the version fetch is reported
+// as rate-limited/transient.
+func TestValidateNPM_VersionEndpointRateLimited(t *testing.T) {
+	ctx := context.Background()
+	mock := newNPMMock(http.StatusTooManyRequests, "", http.StatusOK)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "1.0.0"}
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "rate-limited")
+	assert.NotContains(t, err.Error(), "not found")
+}
+
+// TestValidateNPM_VersionNotFoundProbeUnclassified: the version 404s and the probe
+// returns an unclassifiable status, so the validator falls back to a plain
+// version-not-found message.
+func TestValidateNPM_VersionNotFoundProbeUnclassified(t *testing.T) {
+	ctx := context.Background()
+	mock := newNPMMock(http.StatusNotFound, "", http.StatusTeapot)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "1.0.0"}
+	err := registries.ValidateNPMPackage(ctx, pkg, "io.github.test/demo")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "version '1.0.0' not found")
+	assert.NotContains(t, err.Error(), "exists, but version")
+}
+
+// TestValidateNPM_PositivePathMock: a version response with the matching mcpName validates.
+func TestValidateNPM_PositivePathMock(t *testing.T) {
+	ctx := context.Background()
+	const serverName = "io.github.test/demo"
+	body := fmt.Sprintf(`{"mcpName":%q}`, serverName)
+	mock := newNPMMock(http.StatusOK, body, http.StatusOK)
+	defer mock.Close()
+
+	pkg := model.Package{RegistryType: model.RegistryTypeNPM, RegistryBaseURL: mock.URL, Identifier: "demo-pkg", Version: "1.0.0"}
+	err := registries.ValidateNPMPackage(ctx, pkg, serverName)
+	assert.NoError(t, err, "a version response with the matching mcpName should validate")
+}
+
 func TestValidateNPM_RealPackages(t *testing.T) {
 	ctx := context.Background()
 
@@ -126,6 +299,12 @@ func TestValidateNPM_RealPackages(t *testing.T) {
 
 			err := registries.ValidateNPM(ctx, pkg, tt.serverName)
 
+			// A live 429/5xx from the registry is inconclusive, not a failure
+			// of the case under test.
+			if err != nil && strings.Contains(err.Error(), "retry later") {
+				t.Skipf("transient registry response: %v", err)
+			}
+
 			if tt.expectError {
 				assert.Error(t, err)
 				assert.Contains(t, err.Error(), tt.errorMessage)