fix(e2e): prune smoke-post URLs from sitemap on cms-publish-flow cleanup

e2e/cms-publish-flow.spec.js

@@ -4,6 +4,7 @@ const path = require("node:path");
 const { execFileSync } = require("node:child_process");
 const { test, expect } = require("./base");
 const { captureStep } = require("./manual-capture");
+const { pruneSitemapUrls } = require("./sitemap-prune");
 
 // True end-to-end content loop: drive the live Decap admin to create a new
 // post, rebuild the site, then GET /blog/<slug>/ and assert the post is
@@ -78,6 +79,22 @@ function removeSmokePost() {
   ]) {
     if (fs.existsSync(dir)) fs.rmSync(dir, { recursive: true, force: true });
   }
+  // ...and prune those URLs from the prebuilt `_site/sitemap.xml`. The
+  // in-test jekyllBuild() baked /blog/<slug>/ (and the manufactured
+  // /tags/<tag>/ archive) into the sitemap; deleting the rendered dirs
+  // above leaves those <loc>s advertised but 404-ing. image-alt-text.spec.js
+  // runs in the SAME e2e-admin job, shares this `_site/`, walks the
+  // sitemap, and fails on the orphaned 404 — so keep the sitemap
+  // consistent with what's actually on disk.
+  const sitemap = path.join(REPO_ROOT, "_site", "sitemap.xml");
+  if (fs.existsSync(sitemap)) {
+    const xml = fs.readFileSync(sitemap, "utf8");
+    const cleaned = pruneSitemapUrls(xml, [
+      `/blog/${SMOKE_SLUG}/`,
+      `/tags/${SMOKE_TAG_SLUG}/`,
+    ]);
+    if (cleaned !== xml) fs.writeFileSync(sitemap, cleaned);
+  }
 }
 
 function jekyllBuild() {

e2e/select-specs.js

@@ -297,6 +297,12 @@ const SPEC_RULES = {
     /^_posts\//,
     /^_layouts\/(post|default)\.html$/,
     /^_includes\//,
+    // Cleanup helper that prunes the smoke post's orphaned sitemap URLs.
+    /^e2e\/sitemap-prune\.js$/,
+  ],
+  // Pure-node unit test for the sitemap-prune cleanup helper.
+  "e2e/sitemap-prune.test.js": [
+    /^e2e\/sitemap-prune\.js$/,
   ],
   "e2e/cms-preview-url.spec.js": [
     /^admin\//,

e2e/sitemap-prune.js

@@ -0,0 +1,30 @@
+// Pure helper: remove <url> blocks from a jekyll-sitemap XML string.
+//
+// Why this exists: cms-publish-flow.spec.js creates a smoke post via the
+// local Decap backend, runs `jekyll build` into the SHARED `_site/` the
+// Playwright webServer serves, then cleans up. Its cleanup deletes the
+// rendered `_site/blog/<slug>/` (and the manufactured tag archive) but the
+// built `_site/sitemap.xml` still advertises those <loc>s — so they 404.
+// image-alt-text.spec.js runs in the SAME e2e-admin job, shares that
+// `_site/`, walks the sitemap, and fails on the orphaned 404 ("expected
+// 200 from /blog/e2e-publish-flow-smoke/, got 404"). Pruning the orphaned
+// <url> blocks on cleanup keeps the sitemap consistent with what's on disk.
+//
+// Kept as a pure, exported function so it's unit-testable without booting
+// Jekyll, Decap, or a browser (see sitemap-prune.test.js).
+
+// Remove every <url>…</url> block whose body contains any of `locNeedles`
+// (matched as plain substrings against the block text, which includes the
+// full <loc> URL). Returns the cleaned XML; unmatched input is returned
+// unchanged. Robust to attribute/whitespace variation because it slices on
+// the <url> element boundaries rather than parsing the whole document.
+function pruneSitemapUrls(xml, locNeedles) {
+  if (!xml || !Array.isArray(locNeedles) || locNeedles.length === 0) {
+    return xml;
+  }
+  return xml.replace(/<url>[\s\S]*?<\/url>\s*/g, (block) =>
+    locNeedles.some((needle) => needle && block.includes(needle)) ? "" : block,
+  );
+}
+
+module.exports = { pruneSitemapUrls };

e2e/sitemap-prune.test.js

@@ -0,0 +1,66 @@
+// @lane: local — pure-node unit test for e2e/sitemap-prune.js (no browser)
+const { test, expect } = require("./base");
+const { pruneSitemapUrls } = require("./sitemap-prune");
+
+// jekyll-sitemap emits a flat <urlset> of <url><loc>…</loc>…</url> blocks.
+// This mirrors the shape cms-publish-flow.spec.js's jekyllBuild() produces,
+// including the orphaned smoke-post + manufactured-tag entries its cleanup
+// must prune so image-alt-text.spec.js (shared _site) doesn't 404 on them.
+const SITEMAP = `<?xml version="1.0" encoding="UTF-8"?>
+<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
+<url>
+<loc>https://adamdaniel.ai/blog/introducing-gha-bench/</loc>
+<lastmod>2026-05-12T00:00:00+00:00</lastmod>
+</url>
+<url>
+<loc>https://adamdaniel.ai/blog/e2e-publish-flow-smoke/</loc>
+<lastmod>2026-05-24T00:00:00+00:00</lastmod>
+</url>
+<url>
+<loc>https://adamdaniel.ai/tags/e2e-smoke-flow-tag/</loc>
+</url>
+<url>
+<loc>https://adamdaniel.ai/</loc>
+</url>
+</urlset>
+`;
+
+test.describe("sitemap-prune", () => {
+  test("removes the orphaned smoke-post and tag URLs, keeps the rest", () => {
+    const out = pruneSitemapUrls(SITEMAP, [
+      "/blog/e2e-publish-flow-smoke/",
+      "/tags/e2e-smoke-flow-tag/",
+    ]);
+    expect(out).not.toContain("/blog/e2e-publish-flow-smoke/");
+    expect(out).not.toContain("/tags/e2e-smoke-flow-tag/");
+    // Untouched URLs survive…
+    expect(out).toContain("/blog/introducing-gha-bench/");
+    expect(out).toContain("<loc>https://adamdaniel.ai/</loc>");
+    // …and exactly two <url> blocks were removed (4 → 2).
+    expect((out.match(/<url>/g) || []).length).toBe(2);
+    // Still well-formed: every opening <url> has a matching close.
+    expect((out.match(/<url>/g) || []).length).toBe(
+      (out.match(/<\/url>/g) || []).length,
+    );
+    expect(out).toContain("</urlset>");
+  });
+
+  test("substring match doesn't over-prune a longer path that contains the needle", () => {
+    const xml = `<urlset>
+<url><loc>https://adamdaniel.ai/blog/smoke/</loc></url>
+<url><loc>https://adamdaniel.ai/blog/smoke-test-results/</loc></url>
+</urlset>`;
+    // Needle "/blog/smoke/" has a trailing slash, so it must NOT match
+    // "/blog/smoke-test-results/".
+    const out = pruneSitemapUrls(xml, ["/blog/smoke/"]);
+    expect(out).not.toContain("/blog/smoke/");
+    expect(out).toContain("/blog/smoke-test-results/");
+    expect((out.match(/<url>/g) || []).length).toBe(1);
+  });
+
+  test("no-ops when there are no needles or no match", () => {
+    expect(pruneSitemapUrls(SITEMAP, [])).toBe(SITEMAP);
+    expect(pruneSitemapUrls(SITEMAP, ["/blog/does-not-exist/"])).toBe(SITEMAP);
+    expect(pruneSitemapUrls("", ["/x/"])).toBe("");
+  });
+});