[Bug]: Error when trace is used: apiRequestContext._wrapApiCall: End of central directory record signature not found. Either not a zip file, or file is truncated. (again)

[aczekajski]

Version

1.62.1

Steps to reproduce

In some cases when the test fails and the trace is recorded (I use on-first-retry), the apiRequestContext._wrapApiCall: End of central directory record signature not found. Either not a zip file, or file is truncated. is being thrown during After Hooks > Fixture "context" step.

I am still trying to find out how to consistently reproduce this problem. For now, what I am noticing is that:

• it consistently happens on every failed test when running on Jenkins - doesn't matter if I use sharding or not, happens as well with --workers=2 and --workers=1
• I've seen this error not being throws when only 2 tests within the same test file failed, while all previous cases when it was thrown happened when there were ~3-5 tests failing across 2 different test files.
• I couldn't reproduce it on my local windows PC

In both cases, the tests are run with PW_TEST_CONNECT_WS_ENDPOINT pointing to a docker machine using mcr.microsoft.com/playwright:v1.62.1-noble.

Expected behavior

Traces should work correctly.

Actual behavior

This error is thrown during After Hooks > Fixture "context" step:
(stacktraces obtained by using PWDEBUGIMPL=1 env var)

apiRequestContext._wrapApiCall: End of central directory record signature not found. Either not a zip file, or file is truncated.
    at captureRawStack (/[...]/node_modules/playwright-core/lib/coreBundle.js:9681:17)
    at captureLibraryStackTrace (/[...]/node_modules/playwright-core/lib/coreBundle.js:56883:17)
    at APIRequestContext2._wrapApiCall (/[...]/node_modules/playwright-core/lib/coreBundle.js:57078:28)
    at ArtifactsRecorder._stopTracing (/[...]/node_modules/playwright/lib/index.js:731:24)
    at ArtifactsRecorder.willCloseRequestContext (/[...]/node_modules/playwright/lib/index.js:673:16)
    at Object.runBeforeCloseRequestContext (/[...]/node_modules/playwright/lib/index.js:165:33)
    at Proxy.<anonymous> (/[...]/node_modules/playwright-core/lib/coreBundle.js:62996:35)
    at APIRequestContext2.dispose (/[...]/node_modules/playwright-core/lib/coreBundle.js:58859:37)
    at _BrowserContext.close (/[...]/node_modules/playwright-core/lib/coreBundle.js:62199:28)
    at close (/[...]/node_modules/playwright/lib/index.js:391:23)
    at Object.context (/[...]/node_modules/playwright/lib/index.js:434:13)
    at /[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:190:9
    at Fixture._teardownInternal (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:229:7)
    at /[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:1133:11
    at TimeoutManager.withRunnable (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:403:14)
    at TestInfoImpl._runWithTimeout (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:1131:7)
    at TestInfoImpl._runAsStep (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:1122:7)
    at Fixture.teardown (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:209:11)
    at FixtureRunner.teardownScope (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:276:9)
    at /[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:1667:9
    at TestInfoImpl._runAsStep (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:1122:7)
    at WorkerMain._runTest (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:1652:5)
    at WorkerMain.runTestGroup (/[...]/node_modules/playwright/lib/worker/workerProcessEntry.js:1510:9)
    at process.<anonymous> (/[...]/node_modules/playwright/lib/common/index.js:1997:25)
<inner error>
Error: End of central directory record signature not found. Either not a zip file, or file is truncated.
    at /[...]/node_modules/playwright-core/lib/utilsBundle.js:38081:18
    at Immediate.<anonymous> (/[...]/node_modules/playwright-core/lib/utilsBundle.js:38599:11)
    at processImmediate (node:internal/timers:485:21)
    at process.callbackTrampoline (node:internal/async_hooks:130:17)

In report it looks like this:

Additional context

I will keep updating this ticket as I keep trying to figure out what the core reason is or what under conditions it fails. Debugging is kinda hard so if you have any ideas on how to gather some more diagnostics or how to pinpoint it better, let me know.

Environment

Remote Jenkins host machine (the one running `npx playwright test`):
  System:
    OS: Linux 6.13 Debian GNU/Linux 11 (bullseye) 11 (bullseye)
    CPU: (64) x64 AMD EPYC 7713P 64-Core Processor
    Memory: 72.82 GB / 125.74 GB
    Container: Yes
  Binaries:
    Node: 22.15.1 - /usr/local/bin/node
    Yarn: 1.22.22 - /usr/local/bin/yarn
    npm: 10.9.2 - /usr/local/bin/npm
  Languages:
    Bash: 5.1.4 - /bin/bash
  npmPackages:
    @playwright/test: 1.62.1 => 1.62.1 

Docker machine with the playwright server (the one running `npx playwright run-server`):
  System:
    OS: Linux 6.13 Ubuntu 24.04.4 LTS 24.04.4 LTS (Noble Numbat)
    CPU: (64) x64 AMD EPYC 7713P 64-Core Processor
    Memory: 75.39 GB / 125.74 GB
    Container: Yes
  Binaries:
    Node: 24.18.1 - /usr/bin/node
    Yarn: 1.22.22 - /usr/bin/yarn
    npm: 11.16.0 - /usr/bin/npm
  Languages:
    Bash: 5.2.21 - /usr/bin/bash

My local machine (where the error does not happen):
  System:
    OS: Windows 11 10.0.26200
    CPU: (14) x64 Intel(R) Core(TM) Ultra 7 265U
    Memory: 7.89 GB / 31.43 GB
  Binaries:
    Node: 22.15.1 - C:\nvm4w\nodejs\node.EXE
    npm: 10.9.2 - C:\nvm4w\nodejs\npm.CMD
  Languages:
    Bash: 5.2.37 - C:\Program Files\Git\usr\bin\bash.EXE
  npmPackages:
    @playwright/test: 1.62.1 => 1.62.1

[dcrousso]

this looks like the original error while creating the trace archive is being swallowed, with the truncated archive then failing when it is reopened on the client

#42242 changes this error handling, but it is not included in 1.62.1

could you please try playwright@next on both the Jenkins host and remote Playwright server, making sure both use the same Playwright version, and share the new stack trace with PWDEBUGIMPL=1 if this still reproduces?

[aczekajski]

Devin Rousso (@dcrousso) I tried @next (resulting in 1.63.0-alpha-2026-08-25) and it has become a completely different error occuring under the same ciscumstances as above:

browserContext._wrapApiCall: EACCES: permission denied, mkdir '/home/jenkins'
    at captureRawStack (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright-core/lib/coreBundle.js:8474:17)
    at captureLibraryStackTrace (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright-core/lib/coreBundle.js:57170:17)
    at _BrowserContext._wrapApiCall (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright-core/lib/coreBundle.js:57365:28)
    at ArtifactsRecorder._stopTracing (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/index.js:744:24)
    at /home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/index.js:696:18
    at ArtifactsRecorder.didFinishTest (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/index.js:694:5)
    at Object.utilityFixtures._setupArtifacts.auto (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/index.js:167:5)
    at /home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:190:9
    at Fixture._teardownInternal (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:229:7)
    at /home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:1132:11
    at TimeoutManager.withRunnable (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:403:14)
    at TestInfoImpl._runWithTimeout (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:1130:7)
    at TestInfoImpl._runAsStep (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:1121:7)
    at Fixture.teardown (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:209:11)
    at FixtureRunner.teardownScope (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:276:9)
    at /home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:1675:9
    at TestInfoImpl._runAsStep (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:1121:7)
    at WorkerMain._runTest (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:1660:5)
    at WorkerMain.runTestGroup (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/worker/workerProcessEntry.js:1518:9)
    at process.<anonymous> (/home/jenkins/agent/workspace/root-folder-of-the-repository/node_modules/playwright/lib/common/index.js:1941:25)
<inner error>
Error: EACCES: permission denied, mkdir '/home/jenkins'
    at async Object.mkdir (node:internal/fs/promises:859:10)
    at async SerializedFS._performOperation (/home/ubuntu/.npm/_npx/f62224cd87c307ee/node_modules/playwright-core/lib/coreBundle.js:9473:13)
    at async SerializedFS._performOperations (/home/ubuntu/.npm/_npx/f62224cd87c307ee/node_modules/playwright-core/lib/coreBundle.js:9463:13)

Why is PW trying to do something so many folders up the fs tree, even way above the root of the repository? O_O

Additional info:
I had a bit of a problem running this as there is no 1.63.x version of the docker published so I used v1.62.1-noble and then I run the playwright server inside the container like this:

npx -y playwright@1.63.0-alpha-2026-08-25 install chromium webkit && PWDEBUGIMPL=1 npx -y playwright@1.63.0-alpha-2026-08-25 run-server [...]

(otherwise it would complain about missing browser binaries)
It worked fine on my local windows machine connecting to locally run docker machine.

[aczekajski]

I checked if the tests aren't run from some strange working directory but no, this is the pwd: /home/jenkins/agent/workspace/root-folder-of-the-repository on the client machine (the one that runs npm playwright test).

But now when I look at it, isn't the new error happening on the server side (inside docker)? Why would it otherwise be pointing to playwright files from /home/ubuntu/.npm/_npx/f62224cd87c307ee in the "" while everything else points to actually local installation of playwright. But if that was the case, then "/home/jenkins" would make no sense because how would the server even be aware of a path like this, it does not exist on the server.

[aczekajski]

Btw back in the 1.62.1 version, I added some logging where the yauzl3.open is called in PW's coreBundle.js and here's an example filepath of what file it tries and fails to open: /home/jenkins/agent/workspace/root-folder-of-the-repository/.pw-test-results/shard-1/.playwright-artifacts-3/c161782f13a1261b06fec47e4beb7ea7.zip.tmp.

Also, when fs.statSync(tempFile).size is run on that file right before yauzl3.open or inside the callback, it shows 0 bytes.

Let me try adding some more logging in the @next version as well.

[aczekajski]

I created a patch that adds more logging. Since the new error is coming from inside SerializedFS's mkdir method, making it lose info about the original call place, I passed some stacktraces around to see where is the call causing the error really coming from and here it is:

    Error
        at SerializedFS.mkdir (/tmp/pw/node_modules/playwright-core/lib/coreBundle.js:9400:147)
        at Tracing.start (/tmp/pw/node_modules/playwright-core/lib/coreBundle.js:25604:18)
        at _TracingDispatcher.tracingStart (/tmp/pw/node_modules/playwright-core/lib/coreBundle.js:53992:22)
        at /tmp/pw/node_modules/playwright-core/lib/coreBundle.js:19059:81
        at _ProgressController.run (/tmp/pw/node_modules/playwright-core/lib/coreBundle.js:12295:33)
        at DispatcherConnection.dispatch (/tmp/pw/node_modules/playwright-core/lib/coreBundle.js:19059:44)
        at process.processTicksAndRejections (node:internal/process/task_queues:104:5)

which confirms that the "mkdir" that fails is originally called on the server side. Recognizable because I now install the Playwright into /tmp/pw on docker to be able to apply patch to it. So now the real question is what in the world is going on there that playwright server, running on docker, is trying to mkdir '/home/jenkins', while such path exists only on jenkins which runs npx playwright test and connects to the docker.

Devin Rousso (@dcrousso) Shouldn't this ticket be reopened? Or I should file the findings about the new errors into a new ticket?

[aczekajski]

I also added a log to see what's the actual dir passed to that mkdir command and it seems that the server is trying to call mkdir with this path: /home/jenkins/agent/workspace/root-folder-of-the-repository/.pw-test-results/shard-2/.playwright-artifacts-13/resources which the server should not be aware of (it doesn't even have that repo cloned inside). Is the test runner telling the server to create artifacts under some paths while ignoring that the server has a filesystem of its own? o0