Arch Linux

tropicalia

Member

Registered: 2025-11-13

Posts: 17

[SOLVED] Reproducible status

Hi everyone,

I’ve been exploring the reproducible builds status on Arch Linux (https://reproducible.archlinux.org/) and trying to better understand how discrepancies are handled.

From what I gathered, packages seem to fall into three categories:

1. Reproducible (confirmed by the official infrastructure)
2. Reported as non-reproducible, but reproducible under certain local conditions
3. Clearly non-reproducible

While testing locally (using a clean chroot environment), I found a few packages currently marked as non-reproducible that appear reproducible on my system, including:

atuin 18.13.6-1
bat 0.26.1-2
gamemode 1.8.2-2
libphonenumber 1:9.0.27-2
wezterm 20240203.110809.5046fc22-2

I verified this using diffoscope and did not observe differences in my builds compared to the reference artifacts.

Before drawing any conclusions, I wanted to ask:

* Is it expected that some packages marked as non-reproducible may still build reproducibly under certain environments?
* Could this be due to non-determinism that only appears under specific conditions (e.g., toolchain versions, filesystem ordering, timestamps, etc.)?
* Is there value in reporting these findings, or is the assumption that the official infrastructure results take precedence unless reproducibility is demonstrated across multiple independent environments?

If reporting is useful, is there a preferred way to share these results (e.g., attaching diffoscope reports, referencing BUILDINFO, etc.)?

I’d appreciate any guidance on how to interpret these cases correctly and whether they can help improve the reproducibility status overall.

I have many others that are non-reproducible matching the official report that I have findings that I'd like to contribute but my account was not approved yet. I sent an email a week ago to get it approved. I'm waiting.

Thanks!

Last edited by tropicalia (2026-04-10 18:40:44)

Allan

Pacman

From: Brisbane, AU

Registered: 2007-06-09

Posts: 11,672

Website

Re: [SOLVED] Reproducible status

* Is it expected that some packages marked as non-reproducible may still build reproducibly under certain environments?

These could be non-reproducible if the hostname changes, if the timezone changes, if....

* Could this be due to non-determinism that only appears under specific conditions (e.g., toolchain versions, filesystem ordering, timestamps, etc.)?

The reproducible builds infrastructure installs the same packages for every build, so toolchain versions are the same.  Filesystem ordering, timestamps etc are issues.  Although makepkg does its best to fix ordering issues...

* Is there value in reporting these findings, or is the assumption that the official infrastructure results take precedence unless reproducibility is demonstrated across multiple independent environments?

Probably not.

---

PGP Key: F99FFE0FEAE999BD | I develop pacman - buy me a drink!

tropicalia

Member

Registered: 2025-11-13

Posts: 17

Re: [SOLVED] Reproducible status

Thank you