juliank

zygoon, 10 hours ago to random

A family member has died recently and as a part of shuffling stuff around, I've recovered a thinkpad I used to use.

At the moment it's just wiped with a fresh copy of Windows, but I think I will put Debian on it later. I have not used raw Debian on bare metal in a while, I wonder if I should go with testing or stick to stable.

juliank, 18 hours ago to random

fatal: [localhost]: FAILED! => {"ansible_facts": {"discovered_interpreter_python": "/usr/bin/python3"}, "changed": false, "msg": "Dependency is not satisfiable: libglib2.0-0 (>= 2.37.3)\n"}

ansible's apt module with a deb ignores provides, here libglib2.0-0 is provided by libglib2.0-0t64

Not sure where to report bug and lazy on a holiday, just trying to upgrade my backup laptop.

juliank, 1 day ago to random

Zugzwang (zug = move, zwang=compulsion), such an important word. It says you need to make a move.

You use it in chess other games, or in real life.

You can also joke because Zug also means train that it is a "need to take train".

Mellcolm, 2 days ago to random German

Ich schreibe an der Uni Regensburg aktuell meine Masterarbeit zur Akzeptanz von GenAI in der Softwareentwicklung. Jeder Teilnehmerin bekommt einen Stein in meinem Brett. Gesucht sind Entwickler*innen. https://s2survey.net/genai_acceptance/ Pls. weitersagen! (Dauert keine 10 Minuten, schwöre!)

juliank, 2 days ago to random

Gadgetbridge can't download activities from my Fossil Hybrid anymore, I think it times out because it hasn't synced in months.

juliank, 2 days ago to random

Eek, we can forego all the ordering concerns when ordering clauses because quite frankly we already said "If B is to be installed, A or B is satisfied", so it's fine to queue B before A or B.

(It's not really but for all intents and purposes it depends on luck whether you see A or B, or B first.)

juliank, 2 days ago to random

The new APT solver currently uses a priority queue of outstanding work (unsolved clauses) with dynamic priorities.

However my local state is statically ordered instead, and it's really down to 3 bits:

• is this clause unit? Do it now!
• do earlier group before later
• do mandatory before optional

So we can move the optionality into the groups itself, and then we just need one bucket per group which is just insertion ordered, i.e. an array of linked lists will do fine.

juliank, 3 days ago to random

It's very hard to find stuff about SAT solving, you just get "how to master your SATs"

juliank, 2 days ago to random

I'm sure there's a shortcut for changing a selection to be what you have yanked elsewhere, it's not ergonomical to do

• yanking
• "caw" or similar
• Hit esc
• Press P

juliank, 3 days ago (edited 3 days ago) to random

Just uploaded a new build of @keepassxc to Debian unstable, this adds a separate keepassxc-minimal package, and a transitional package that depends on keepassxc-full | keepassxc-minimal; so people upgrading (or new installing) will get the full experience again.

This is still 2.7.7, mind you. 2.7.8 I need to go look at some day when I'm not killing myself with apt solver work!

juliank, 2 days ago to random

I don't know - I test my lung function with singing "running low" in hallowed be thy name.

juliank, 3 days ago to random

Wild Euro plate today:

Ratatouille (French)
Fries (Belgian)
Aioli (Spanish)

Tons of finest extra virgin olive oil from Kalamata, Greece.

juliank, 3 days ago to random

See if you want to eat ratatouille with pasta, orecchiette would be a good choice.

juliank, 3 days ago to random

Ratatouille, or: olive oil with vegetables

juliank, 3 days ago to random

Unable to install x:

x=1 depends Y
y=1 depends Z
z=1 Conflicts alpha

but:

foo=1 is manually installed.
foo=1 Depends bar
alpha=1 Provides bar

juliank, 3 days ago to random

The new branch on the top is much neater. All the dynamic "choose smaller or groups before larger or groups" business is gone, as well the complex logic for renamed packages.

Renamed packages are now handled because we resolve dependencies involving only new packages first, and dependencies involving obsolete packages (no longer in the repository or not built in source candidate anymore) last (well after any other dependencies, not after install requests).

juliank, 3 days ago to random

This is essentially the same data set (different runs), on the left record with perf, on the right with gperftools profiler; then (perf converted to pprof with https://github.com/google/perf_data_converter and) rendered with pprof.

juliank, 3 days ago to random

Solver profile graph from https://github.com/google/pprof, with relative percentage (restricted to the solver).

For the complex 1418 64-bit time_t upgrade scenario early in noble.

juliank, 3 days ago to random

We have hilarious inefficiencies in the solver right now:

• We potentially discover the same package's dependencies multiple times

• We add clauses for dependencies shared between across all versions for both the package and the selected version (we don't keep track of what we merged).

None of these are performance critical yet.