[Bug 2029268] Re: Do not consider two versions with differing SHA256 to be the same

[Andreas Hasenack]

I checked previous SRUs, and we do rely on the automated tests for such
changes, as it can be difficult to fabricate something "in real life"
that would remain valid for the time we need it to be. I'm glad tests
are being added.

This being a native package, and imported into the archive without rich
history, it's a bit hard to follow what was described in d/changelog and
the actual changes for a package I'm unfamiliar with, but I checked with
the upstream branch in salsa, and it looks ok.

The po/* changes, and doc changes, are just about the version, so should
be no issue for translators.

The same set of changes is applied to jammy and lunar, and in devel.

** Changed in: apt (Ubuntu Lunar)
       Status: In Progress => Fix Committed

** Tags added: verification-needed verification-needed-lunar

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to apt in Ubuntu.
https://bugs.launchpad.net/bugs/2029268

Title:
  Do not consider two versions with differing SHA256 to be the same

Status in apt package in Ubuntu:
  Fix Committed
Status in apt source package in Jammy:
  Fix Committed
Status in apt source package in Lunar:
  Fix Committed
Status in apt source package in Mantic:
  Fix Committed

Bug description:
  [Impact]
  APT sometimes deduplicates two debs into the same version object even if they have different SHA256 field values, causing download to fail later if one the sources also defines SHA512 (or MD5 or SHA1).

  This is a problem for example, if you rebuild in a PPA because PPAs do
  not have SHA512 enabled but the priamary archive does.

  Repositories are not required to have SHA256, so this does nothing if
  we do not have SHA256 for both .deb.

  [Test plan]
  An automated test is included in apt's extensive autopkgtest regression test suite. Successful pass of autopkgtest is the goal.

  [Where problems could occur]
  In terms of regressions it seems unlikely, because we compare the SHA256 only if we previously would have considered them the same version to reject them if they differ.

  But of course there could be the usual unsafe memory bugs.

  In a future this will bite us when we migrated to SHA3 and want to
  drop SHA256, just like we cannot seem to drop MD5 now.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/apt/+bug/2029268/+subscriptions