Skip to content

Debian/Xen microsprint, Cambridge, 11th September 2018

Goals for the day:

  • Get an ACCEPTED email about Xen 4.11 in Debian Unstable
  • Choose some interesting discussion topics involving upstream developers
  • Have fun! :)

We have limited time, so the order of things is:

  • Fix show stoppers / deal breakers first.
  • Look at other issues and either fix them if it takes 5 minutes, or document all the what / why / how to fix and postpone the actual work, since it can be done individually.
  • Have interesting discussions about things.

Open issues with 4.11 packaging

  • Ian: Remove archaic packaging things, end up with a smaller, readable debian/rules

In the salsa list, there's all the lintian issues. Solve them, or document why they don't matter, or how to solve them.

These are the show stoppers:

While doing the above, I ended up adding a stable-4.11-dfsg branch, which removes the dfsg offending file from the upstream code, and added the tag upstream/4.11.1_pre.20180911.5acdd26fdc+dfsg on it. We added the date also, to give a user a bit more feeling about how old it is. If there would be an extra upstream version change today, we'd add a +1 just before the git commit id.

Other important ones:

  • The libxentoolcore story is the next important one, since it's about breaking package upgrades. -> Fixed by package rewrite.
  • Now look at the other issues tagged lintian. -> Almost everything is irrelevant after package rewrite, I deleted them.
  • Ian says: "The patch to reintroduce xs_restrict should go upstream.". -> in progress

Meanwhile, practice using git-debrebase.

Upload!

At this point, get everything building, smoke test and upload to Debian, yay.

The rest from the salsa list

Either fix these right away, or move them to the bts with extra ideas how to fix them:

Things from the bts

A few interesting ones, maybe timebox them when discussing:

  • libxen, libxen-dev: install libxenvchan library -> where to put this? maybe related to the xentoolcore story? libxen-'basic' or similar package? -> Fixed by package rewrite.
  • gnttab max frames... what will we finally do with this? -> idea: reproduce, use log_once to improve upstream error reporting when it happens, send patch upstream.
  • iptables antispoofing rules... the current thing is a bit misdirected and non-functional. Interesting discussion is... what kind of anti-spoofing functionality would be good to have out of the box (apart from how to implement this, which is step 2) -> noop now
  • Crashing Windows VM... interesting thing is: how do we handle these kind of reports, where the user himself can only test -> Yes, just be honest and tell the user we cannot help them.
  • Grub menu and locales... related: GRUB i10n -> translations are in the grub package, not in xen. I probably tested with a language that has no translation, and couldn't reproduce. Look at this again. Ian says: patch looks fine.

And another patch from Ubuntu that I didn't merge yet because I don't know the specifics, let's look at it:

Things in discussion / brainstorm phase

Moar things

  • Cosmetics: squash together the whole abiname/prefix patches into two commits. Irrelevant after package refactoring.
  • geub hvc0 thing (note to self)

Moar triaging

There's some accumulated cruft, but also interesting things in our bts list. If there's any time left, we could look over them and see if there's any low hanging fruit that can be solved or closed.

For remaining bugs, it makes sense to contact the submitter and ask them to confirm they still are experiencing the problem, and ask them to reproduce in a Debian Buster setup. In some cases, it's very specific, and the bug submitter is the only one who can re-test and provide more info.

Edited by Hans van Kranenburg