Gajim - 2025-09-21

how do I get pyright to run locally like in runs in CI, ie, only on diffs from the last X commits?

pyright does not run on diffs

it runs over the whole code base, and you need to install it, https://github.com/microsoft/pyright/blob/main/README.md

for real? when I run it locally, it throws A LOT of errors, and in CI it looks like only the MR diff is processed, I guess it's on the whole modified files list

lol no > pyright (git diff --name-only) [...] 1049 errors, 0 warnings, 0 information I'll let it run in CI, there is something I don't understand about how it's launched in the gitlab pipeline I guess

you simply have not the right dependencies installed

what is the error?

a lot of them, but related to gtk stuff indeed

eg /home/nicoco/src/perso/xmpp/gajim.clean/gajim/gtk/avatar.py:709:47 - error: Type of "get_height" is unknown (reportUnknownMemberType)

Oh I got it, I thought `uv sync --dev` would install [project.optional-dependencies] but this field actually defines an "extra", so I need to `uv sync --all-extras` to get pygobject-stubs to install. Yay! ✏

I wanted to add uv instructions, and use it for Windows builds, as it's way faster than pip. But I encountered a build error along the way back when I tried. Might try again soon :)

uv installs things in virtual environments

That's not what we want when building the windows installer

Not sure if it's even possible to not install in a virtual env

it is a bit weird to define dev dependencies as an "extra" I think, but gajim is not published to PyPI and there is no standard way of defining "dev dependencies" anyway, so this works I guess

uv can install "system packages" too, although I have no idea about whether that actually is something in windows

nicoco: why would this be weird, how would it be possible differently?

poetry, uv (and probably hatch and others) have their own syntax for that

uv uses: ```toml [dependency-groups] dev = [ "coverage>=7.6.10", "emoji>=2.14.0", [...] ``` for instance

So what's weird? Still is defined as extra

it's a bit weird to use "extras" because this usually define runtime dependency for optional behaviour, to do stuff like `pip install some-package[some-extra-feature-that-requires-more-runtime-dependency]`

> So what's weird? Still is defined as extra no, extra is a standard thing of pyproject.toml, the field is `[project.optional-dependencies]`, not `[dependency-groups]` ↺

[project.optional-dependencies] has consequence when publishing to PyPI, for instance

Why do you think extras should be used only for runtime requirements? ✏

it's not me

who think that

but people writing project management tools like uv and poetry

but I guess you're right, https://packaging.python.org/en/latest/specifications/dependency-specifiers/#extras does not mention anything about it being runtime only

the example even includes a "tests" extra

Surprise 😄

Well, the people behind uv and poetry disagree, and I think it's fair to assume they have valid reasons.

I know I have a use-case for slidge

I use some very custom sphinx extensions for document, which I don't particularly want to publish to PyPI.

Stuff not published on PyPI is forbidden in [project.optional-dependencies] and prevents publishing on PyPI.

The same with extras

But uv lets you define a custom pypi index, or even "git dependencies", and you can use those in [dependency-groups]

> The same with extras Yes, precisely my point, with uv "dependency groups" you can put stuff that is not necessarily on PyPI. ↺

Ok but how is that relevant, do we need something that is not on pypi?

absolutely not, lol

just discussing is all ;)

well actually gajim does define "git dependencies"

so the current pyproject.toml would prevent publishing gajim on pypi I think. but gajim is not published to pypi.org anyway, is it?

No

then you're right, we don't care. I'm happy to have slidge published on PyPI.org (which makes a lot of sense for a lib), and still be able to use some non-pypi.org dependencies for development.

Makes not much sense as it has non python dependencies ✏

> Makes not much sense as it has non python dependencies some non-python stuff is packaged on pypi too these days. ever used CUDA? :) ↺

We also need to install files into system locations

No tool can do this to my knowledge

I don't know and I decided not to spend more time on that, the gtk, gobject, etc. dependencies are a bit arcane to me.

Which is why I basically decided a few months ago that I'll get back to contributing to gajim whenever I update my computer to debian trixie, which I did a few days ago :)

yeah its a bit sad that we cannot have a dev env thats independent of the system

it will always stay compatible with latest ubuntu

Anyway, I think 2 of my MRs are ready (missing context for non-joined MUCs, use affiliaations to compute MUC name). For the 'displayed marker in MUCs' I think we do not want to store all markers forever so we probably need to do some sort of purging, but I'm not sure where to place it.

About the purging, the first idea that comes to mind is "just store the last received marker by occupant" but I think it's a bit more complex than that. The rule should be "store the lastest marker, but only if it refers to a newer message than the marker we have in store for this occupant", but I'm not sure where to implement that.

I was wrong earlier, [dependency-groups] is not uv-specific but standardized too. (it does not change anything for gajim but still, better to correct oneself when one said wrong things, right?) https://packaging.python.org/en/latest/specifications/dependency-groups/#dependency-groups

thanks, yes we should then switch to dependency-groups even if it makes no difference for us, seems this key was introduced specifically for development tools

about the markers, it is our concept that we store messages without looking at the database beforehand

so building logic such as "store this only if not something different is already stored" is problematic. Also it makes things much more complex because you have a bunch of assmumptions in that logic, e.g. that messages are retrieved in a specific order or that something is stored already at a specific moment

our concept is that we store messages and metadata without any checks.

you apply your logic and checks when you retrieve the data, then you can apply whatever logic is needed

lovetox pushed 2 commits to branch python-nbxmpp/master imprv: Log details about invalid roster items - https://dev.gajim.org/gajim/python-nbxmpp/-/commit/2d028d07e5f98a538a8db06ffa8c9fa073814c12 imprv: Log details about invalid MUC affiliation items - https://dev.gajim.org/gajim/python-nbxmpp/-/commit/bed88d4f49daee03d93ba1b77eb332adb97cc565

> our concept is that we store messages and metadata without any checks. The MR does that so it's more ready than I thought it was. =) I can't help but wonder if we shouldn't add some sort of cronlike autopurge every X days or hours, and only keep the most X recent markers by (MUC, occupant) pairs, but that's an optimisation that can come later, if necessary, I guess. ↺

hello

Does prosody server support SCRAM-SHA1-PLUS ?

yes, i see people logging to my prosody with SCRAM-SHA-1-PLUS

speaking of this, does gajim support channel binding?

No

where do I find my xmpp address

Click the avatar in the left bottom

lovetox: I see the talk with mebs, Gajim does not work with Channel Binding (-PLUS variants)? So, I need to change Gajim here: - https://github.com/scram-sasl/info/issues/1 - https://wiki.xmpp.org/web/SASL_Authentication_and_SCRAM

Yes Gajim does not support channel binding