From 26707275ffe2214d7d6443d804cdd120cfd59423 Mon Sep 17 00:00:00 2001
From: meetsu <96637888+klNuno@users.noreply.github.com>
Date: Mon, 19 Oct 2026 02:38:16 +0200
Subject: [PATCH] fix(ci): retry a quarantined test that fails in a run of
 several files

---
 scripts/ci/flaky.test.ts | 13 +++++++++++++
 scripts/ci/flaky.ts      |  3 ++-
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/scripts/ci/flaky.test.ts b/scripts/ci/flaky.test.ts
index b37dfacca..5748bcd84 100644
--- a/scripts/ci/flaky.test.ts
+++ b/scripts/ci/flaky.test.ts
@@ -19,6 +19,19 @@ test('failed tests are named only when every counted failure has a name', () =>
   expect(failedTests('(fail) a\n 1 fail\n(fail) b\n 1 fail\n')).toEqual(['a', 'b']);
 });
 
+test('a run of several files names each failure once, not again in its closing recap', () => {
+  // bun test over several files lists every skip and failure again before the counts.
+  const several = [
+    '(fail) relays reads [4246.43ms]', '(pass) other [1.00ms]', '',
+    '1 tests skipped:', '(skip) killed outright', '', '',
+    '1 tests failed:', '(fail) relays reads [4246.43ms]', '',
+    ' 111 pass', ' 1 skip', ' 1 fail', 'Ran 113 tests across 39 files. [447.52s]', ''
+  ].join('\r\n');
+  expect(failedTests(several)).toEqual(['relays reads']);
+  const two = '(fail) a [1ms]\n(fail) b [1ms]\n\n2 tests failed:\n(fail) a [1ms]\n(fail) b [1ms]\n\n 2 fail\n';
+  expect(failedTests(two)).toEqual(['a', 'b']);
+});
+
 test('a retry needs every failure quarantined in a file that ran', () => {
   const list = [entry('tests/e2e/a.test.ts', 'one'), entry('tests/e2e/b.test.ts', 'two')];
   const absolute = join(ROOT, 'tests', 'e2e', 'a.test.ts');
diff --git a/scripts/ci/flaky.ts b/scripts/ci/flaky.ts
index b7a882c07..d096b5a53 100644
--- a/scripts/ci/flaky.ts
+++ b/scripts/ci/flaky.ts
@@ -14,7 +14,8 @@ export const QUARANTINE: Quarantined[] = JSON.parse(readFileSync(resolve(import.
 
 /** Names of the failed tests in bun test output, or undefined unless every counted failure has one. */
 export function failedTests(output: string): string[] | undefined {
-  const clean = output.replace(/\x1b\[[0-9;]*m/g, '');
+  // Over several files, bun lists every failure again under "N tests failed:" before the counts.
+  const clean = output.replace(/\x1b\[[0-9;]*m/g, '').replace(/^\d+ tests? failed:\r?\n(?:\(fail\) .*\r?\n)*/gm, '');
   const names = [...clean.matchAll(/^\(fail\) (.+?)(?: \[[\d.]+m?s\])?\r?$/gm)].map(match => match[1]!);
   const count = [...clean.matchAll(/^\s*(\d+) fail\r?$/gm)].reduce((sum, match) => sum + Number(match[1]), 0);
   return names.length > 0 && names.length === count ? names : undefined;