Repository navigation
postject sometimes cannot inject the SEA blob on Windows, resulting in flaky test #49630
Description
Activity
My Windows setup is not working right now. I can try to take a look in a couple of days.
It can also fail on Linux during loading, I suspect we need https://chromium-review.googlesource.com/c/v8/v8/+/4608330 to fix it, but that seems separate from the postject failure on Windows
Not sure why but this has stopped showing up in the CI..
Reacted by Darshan Sen- added a commit that references this issue
on Oct 18, 2023 - added a commit that references this issue
on Oct 23, 2023 - added a commit that references this issue
on Nov 11, 2023 I am not sure why they are still showing up in reliability reports after #50223 but they are still flaky, and it appears that other SEA tests are flaky too probably due to the same reason nodejs/reliability#718 - SIGSEGV when executing the SEA, on RHEL, Windows and macOS. Maybe we should open a separate issue because the original issue happened during injection.
This also shows up for the asset test on Windows: https://ci.nodejs.org/job/node-test-binary-windows-js-suites/26136/RUN_SUBSET=2,nodes=win2016-COMPILED_BY-vs2022-x86/console. I suspect that postject may have trouble injecting slightly bigger files under certain circumstances.
14:23:39 �[36mStart injection of NODE_SEA_BLOB in C:\workspace\node-test-binary-windows-js-suites\node\test\.tmp.960\sea.exe...�[0m 14:23:39 �[31mError: Aborted(). Build with -sASSERTIONS for more info.�[0mWe could probably investigate building it with ASSERTIONS.
- changed the title
[-]test-single-executable-application-snapshot(-and-code-cache) is flaky[/-][+]postject sometimes cannot inject the SEA blob on Windows, resulting in flaky test[/+]on Feb 26, 2024 I wonder if we just skip the test if any of the following failed:
- Copy
- Inject
- Code sign
In those cases, the failures are unlikely to be fixable by Node.js core anyway...though it might still be useful to know when it fails 100% which indicates that something is wrong with our workflow. Not sure how we can know that in single tests though.
- added a commit that references this issue
on Mar 1, 2024 - added a commit that references this issue
on Mar 7, 2024 This seems to have disappeared. Closing.
Refs: nodejs/reliability#663
And from a recent run that includes #49543 https://ci.nodejs.org/job/node-test-binary-windows-js-suites/23086/RUN_SUBSET=1,nodes=win2016-COMPILED_BY-vs2022-x86/console
It looks like sometimes postject is unable to inject the blob? cc @RaisinTen