[Rpm-maint] [rpm-software-management/rpm] Can we guarantee that `rpmtsOrder()` is deterministic? (Discussion #4052)

[Jonathan Lebon]

In https://github.com/rpm-software-management/dnf5/pull/2522, we're discussing whether DNF should sort packages when `SOURCE_DATE_EPOCH` is set before installation to ensure reproducibility even if input package order slightly changes. One concern though is that all this relies on the fact that `rpmtsOrder()` is currently deterministic; i.e. for the same input order, it'll always result in the same output order post-tsort.

That _seems_ like a reasonable assumption to make; AFAICT we're not purposely adding randomness in the reordering process. But the ask is stronger than this: can we make guarantees that `rpmtsOrder()` will remain deterministic? If so, can we document this?

(Note to be clear: this puts no constraints on what that final order actually is nor on whether it can change across rpm versions. We're scoping this to a fixed build environment.)

-- 
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/discussions/4052
You are receiving this because you are subscribed to this thread.

Message ID: <rpm-software-management/rpm/repo-discussions/4052 at github.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rpm.org/pipermail/rpm-maint/attachments/20251121/047793b9/attachment.htm>