Skip to content

Conversation

@alexhulbert
Copy link
Member

The GCP image tarball wasn't producing reproducible hashes between builds on this branch, but worked on main. Additionally, the tarballs we're getting measured properly.

This was due to several issues:

  • FAT ESP partition file/dir timestamps
  • Timestamps in tarball

I did touch -d "@$SOURCE_DATE_EPOCH" on all the ESP files/dirs, added --invariant to SYSTEMD_REPART_MKFS_OPTIONS_VFAT and added --mtime="@$SOURCE_DATE_EPOCH to the tar command and that made the systemd-repart builds reproducible.

@niccoloraspa niccoloraspa requested a review from bakhtin January 27, 2026 17:19
@alexhulbert alexhulbert force-pushed the ah/fix-buildernet-targz branch from 810aa97 to 4433bbc Compare January 28, 2026 23:49
@alexhulbert alexhulbert changed the base branch from trunk/buildernet to ab-gcp-fixes January 28, 2026 23:49
@alexhulbert
Copy link
Member Author

I ended up having to use the approach from the main branch. systemd-repart doesn't let you hard code the first usable sector in the GPT metadata. Once the branches are merged, I'll switch back to using your file and edit the measurement repo to use the first usable sector chosen by systemd-repart.

@bakhtin bakhtin merged commit 45fcb50 into ab-gcp-fixes Jan 29, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants