Gajim - 2023-10-02

hmm, I might be wrong, but I don't think that is a feature in XMPP? I also find that somewhat privacy invasive.

> hmm, I might be wrong, but I don't think that is a feature in XMPP? I also find that somewhat privacy invasive. it has on cheogram (conversations fork)

more privacy-invasive than presence in general?

not by much, i'd think

still it's of arguable usefulness

Polixgen: it is standardized (Kris: XEP-0319, it's even "stable", wow) and implemented. Go to your contact list, it's in the tooltip over contacts. Ofc, your contact must use a client that supports this feature, and they must have it enabled. I think it should be more easily, my WIP attempts: https://dev.gajim.org/gajim/gajim/-/merge_requests/977 and https://dev.gajim.org/gajim/gajim/-/merge_requests/976 ;)

XML Console shows this: ``` <presence xmlns="jabber:client" id="" xml:lang="en" from="slixfeed@" to=""> <show>chat</show> <status>News items: 9497</status> <x xmlns="vcard-temp:x:update"> <photo /> </x> </presence> ``` But Gajim doesn't show the status message: "News items: 9497".

tmux

sorry, wrong window

https://xmpp.pimux.de/file_share/473f9046-ca4b-4daf-a990-ba52279f70bb/gajim_status_slixfeed.png

(Raleigh?)

ClassicOS98 Theme

Lightning Bjornsson, https://github.com/ClassicOS-Themes/ClassicOS-2000-Theme

real

Gentlemen, I need help with the status message please.

Only the blokes? S••t. Otherwise I'd ask what you mean and try to figure out how to get to the bottom of it.

classic ellenor

your cert is expired rn

lovetox, it appears that status messages are shown only in MUC list/roster.

> umu a écrit : > classic ellenor I know someone by that name!

> umu a écrit : > your cert is expired rn you should be expired rn

I am not expiring fast enuf.

🙄

I am having trouble with multi-device syncing

Messages read on one device doesn't show as read on gajim. What must be the problem ?

Kevino: in public rooms that's expected as XMPP currently lacks a standard for read status in public rooms.

no i guess its also happening in private chats

Private groups or 1:1 chats?

1:1

Are the contacts offline or online?

let me check again

Because it's expected for offline contacts. Does your server support carbons and MAM? (if not, change server, it's needed for a pleasant multi device experience…) ✏

i checked with my own accounts all online , but on gajim it stays unread

yes both support message carbons and mam

What other app do you use? Are you sure it sends "read markers"?

ohh 😅️

i think i turned off message read status for contacts

that might be the problem ?

Yeah this is a limitation of XMPP right now. Read state relies on sending read markers to contact (or groups), which is not ideal at all.

ok thanks

A draft spec for improving this was started a while ago: https://xmpp.org/extensions/inbox/inbox.html - but it's only a draft for now

And I know that gajim's maintainer does not like the current situation of read status relying on sending read markers to the recipient, so gajim will probably implement something better some day ;-)

> Yeah this is a limitation of XMPP right now. Read state relies on sending read markers to contact (or groups), which is not ideal at all. TIL, yeah that isn't ideal. ↺

Kris: tigase folks have come up with a hack relying on "self MUC-mediated read markers" for public rooms, but it's not standardized and relies on MUCs enabling PMs. (and it only works in the tigase ecosystem)

Yeah, these hacks are not ideal either

yeah

Raaaaaahhh I wrote a long MR description in gajim's gitlab and it was "detected as spam", and "previous" in the browser does not recover the content. :rage:

😨

I guess it's related to the issues of this weekend, they probably were under attack and had to mitigate with aggressive spam measure protections, I also had trouble commenting on issues yesterday…

Why dtatus messages are shown in Chat but are absent from Roster and Profile dialog?

Schimon: your question has been heared. Please be patient.

cal0pteryx, I just made it clear that status messages are supported and read by Gajim. Previously, I implied that status messages are shown only in MUC.

Schimon: the contact details dialog does not show the presence status message. Currently, AFAIK, the presence status message is only shown in the "Contact list" tooltip.

But in the next version it'll be in the "chat banner", cf https://dev.gajim.org/gajim/gajim/-/commit/442328c20ca0bf15914b9aa1387d87271260c998

(and maybe elsewhere too: <https://dev.gajim.org/gajim/gajim/-/merge_requests/977> and <https://dev.gajim.org/gajim/gajim/-/merge_requests/976>)

Great. Thank you, nicoco.

Staus messages are quite useful when interacting with bots like MAXS which shows battery status and RSS bot which shows how many news items are available.

is it only me, or gajim is unable to send images with size below 5KB?

mrdoctorwho: anything in the logs?

meson, could you remind me please how to start gajim in debug mode? Looks like the -v option is ignored

> Polixgen: it is standardized (Kris: XEP-0319, it's even "stable", wow) and implemented. Go to your contact list, it's in the tooltip over contacts. Ofc, your contact must use a client that supports this feature, and they must have it enabled. I think it should be more easily, my WIP attempts: https://dev.gajim.org/gajim/gajim/-/merge_requests/977 and https://dev.gajim.org/gajim/gajim/-/merge_requests/976 ;) nicoco, where is the tooltip?

Bottom left icon with your own avatar, you see your contact list, leave the mouse cursor on a contact's row for one 1 sec, it should appear.

Polixgen: it is buried but it should be unburied in the next release. Note that unlike some other messengers, this is not an indication of when this contact has last sent a message: it is something that their client optionally sends. Gajim will display whatever their client has sent. Also, not a lot of clients actually send this payload, so you will not see it on a lot of contacts, depending on what they use.

mrdoctorwho: linux or Windows?

meson, linux

Mhm, I only know about "-v".

is there any plans to make gajim a bit faster? or its a python limitation that if you switch the chats you have 1.5sec lags on ryzen 4700?

its not a trolling.

+

ann: do you have participants list open?

It should be much faster if it is closed

> ann: do you have participants list open? I'm switching between private conversations, not the MUCs. ↺

by the way I'm using XMPP as a main messenger and there are 100k+ messages in MAM, that always trying to re-sync to achieve consistency

that slowdowns started to happen in Gajim 1.7+ and in 1.8 its just got worse

xmpp server is also selfhosted so no limits to re-sync archived messages unlike public instances, so that may be the case

Is it really 1.5sec?

I'll make a vod

ann: you could profile gajim and look where it consumes cpu time on switching

I also have a lot of messages and I agree that switching chats, even 1:1, is not "snappy". But it's certainly below 1.5sec.

no. it's more likely 0.9 sec.

Gajim uses only 54%WCPU when switching chats

https://im.ebala.net:5821/upload/pjO28LEJHJqTQRKj5cYevuCT/4FVEVxEfTaifqRncnmEsXQ.mp4

so that's not likely python singlethreading bottleneck

ann, you could use `perf top -p $(pidof gajim)` to see which C function calls it’s doing, make sure you have debug symbols installed.

You could also start Gajim with --cprofile to see which Python function calls it’s doing.

I'll do that, thank you.

A few years ago it was way worse but the devs already did impressive work on gajim's "snappiness".

Gajim 1.4 (or whatever version it was on oldstable Debian) was blazingly fast by the way

but idk is it regression in Gajim or in Gtk3

its the opposite to me so thats definetily weird

from the recording, the first switch took about 1 second, the 2nd 0.7 s and the last one 170 ms.

and according to the changelog it should be faster

yes as I told it was 0.9 sec. still it is veey slow

missed a bit with 1.5, but still

an, alternatively you can use `snakeviz` to analye profile gajim. See https://jiffyclub.github.io/snakeviz/ ✏

an, alternatively you can use `snakeviz` to analyie the profile. See https://jiffyclub.github.io/snakeviz/ ✏

an, alternatively you can use `snakeviz` to analyze the profile. See https://jiffyclub.github.io/snakeviz/ ✏

Is the cprofile file stored anywhere?

Or do we just get the pstats output?

I don't know much about profiling tools (to say the least), and it's not as slow for me as it is for ann (I am on a Ryzen 9350X) but I tried to profile a gajim session where I frantically switch between tabs, with my main gajim profile (where I have a huge history of messages in a lot of MUCs).

https://upload.slidge.im:5281/file_share/2VVQ5eOUTaRm_iCLqxfgBehb/Screenshot_20231002_142651.png

nicoco, looks similar for me

I'm not sure I'm reading that right… Does this mean `gajim.gtk.conversations.view.ConversationView._insert_message()` is where a potential optimization would be useful? In the body of the func itself or maybe call it less often, or maybe delegate to something done in the background…?

I'll try to profile a things and will return with results

https://share.conversations.im/meeson_/qOX0YHkurOBJT2rr/Screenshot_20231002_143634.png

nicoco, that’s one location where this would be useful yes, it would gain 2s of CPU time over the entire run time of your profile.

Here when switching tabs a few times, gajim/gtk/conversation/view.py:259(_reset) is the heaviest function call, it’s only called 15 times but takes 28ms per call.

Just removing all rows from the list box.

I think you’d gain a lot by moving to GTK 4 and using the list model instead.

Probably gain more from moving to a compiled language. ;)

VA1DER, during a tab switch, most of the time seems to be spent removing and readding widgets, no language change will fix anything there.

GTK is already written in C, which is reasonably efficient.

An efficient language can't fix an inefficient algorithm:)

Does Gajim support Chinese typing?

Tab switches seem to be reasonably performant. Going from here to Prosody's chat room is an eyeblink. Does it really matter if it's 100ms or 400ms? I wouldn't think that is worth the work and pain involved in a major widgetset change.

VA1DER, a user before you reported it would take as much as 1s.

This isn’t an eyeblink.

1) Is that an objective or subjective "one second"? Your numbers seem to suggest a switch takes ~420ms. 2) When there are visual cues telling the user that the switch is in progress (the destination tab remaining highlighted even when the mouse moves off it) then the user knows it's working and one second isn't an unreasonable delay. 3) With almost 100% certainty the user will be running on a multi-core system where one second of high cpu usage on one core is unlikely to affect other processes.

Hi, I'm currently encountering an error in my client because another user (probably a different client) has an OMEMO `signedPreKeyPublic` with a `signedPreKeyId` of zero (0) which trips up the code over here: https://dev.gajim.org/gajim/omemo-dr/-/blob/a953803d64c80146e818cf62d15a205f0dc206cc/src/omemo_dr/state/prekeybundle.py#L57 `eu.siacs.conversations.axolotl` is used here, which I assume is XEP-0384 0.3.0. Reading that version I don't see any limitation that would prohibit the use of 0 as an ID, while the current version of the standard, using different names ( `signedPreKeyPublic` vs. `spk` either way), does explicitly disallow 0, while also using 0 in the following example (https://xmpp.org/extensions/xep-0384.html#bundles) which is rather confusing. To keep this short, it seems that there are implementations that have an id=0 for the axolotl's `signedPreKeyPublic`, I just wanted to throw this information *somewhere*, just in case this was not known. At the very least I haven't found much in terms of the error message in the link to the corresponding code. I can try and figure out which client used that id if it's important and/or someone wants to follow up on that.

Katze, yes it would be interesting which client used this

i need to look if this leads anywhere to problems allowing 0

> Katze, yes it would be interesting which client used this I'll try and figure it out, the peer sadly uses like five different clients so it's gonna be fun. I'll get back to you.

yeah seems hard to do actually

in the protocol only device ids are used, but they are usually never exposed somewhere in the GUI

when does this error show in the logs? ✏

when you try to send a message?

does the contact also use Gajim?

to fix it, the contact could use gajim advanced pep config dialog, to delete the bundle

and then hopefully the client will publish something with a higher ID

> yeah seems hard to do actually oh, don't worry, I'll get through this and I'll have fun figuring out things with device IDs and diving into XML. I wanted to acquaint myself with the XMPP OMEMO impls for a while. > when does this error show in the logs? on startup of gajim. When I just run it from the CLI without parameters it's a single error that's being logged, if I run it with verbose flags I get a *lot* of additional information before and after, including the full response that fails the check. I'll see if I can grab the query that causes it real quick.

hm, actually you can figure it out i think

it is a device for which you dont have a fingerprint

because gajim can not parse the bundle, you cant see the key in the trust dialog

so just ask you contact for all his device fingerprints

The error: ```text 10/02/23 14:52:06 (E) gajim.c.m.omemo (benary.org) Building session failed: Signed pre key id out of range ``` ```text 10/02/23 14:52:05 (I) omemo_dr.session_manager (benary.org) Saved devices for [person] 10/02/23 14:52:05 (I) omemo_dr.session_manager (benary.org) Missing device sessions for [person]: [1984947357] 10/02/23 14:52:05 (I) nbxmpp.m.omemo (benary.org) request_bundle(jid=[person], device_id=1984947357) 10/02/23 14:52:05 (I) nbxmpp.m.pubsub (benary.org) request_items(node=eu.siacs.conversations.axolotl.bundles:1984947357, max_items=1, jid=[person]) ```

or simply write him a message, and ask him on which device he didnt receive the message

thats the one then

you cannot figure it out from the logs, because device ids are never linked to presence of devices

but yeah i will see that omemo lib accepts 0

its the lesser evil i think

so basically all these gateways to messengers break e2e encryption

> Is there anything wrong with this stanza? > ``` > <message type="groupchat" id="1758461952" from="-608413854@zg.94.177.207.80/D/K (1382500964)" to="bqdiscord@94.177.207.80"><subject>1758461952 | D/K (1382500964) | 18 Sep 2023 04:51:17 | хрк</subject><delay xmlns="urn:xmpp:delay" from="-608413854@zg.94.177.207.80" stamp="2023-09-18T04:51:17Z"></delay><x xmlns="jabber:x:delay" from="-608413854@zg.94.177.207.80" stamp="20230918T04:51:17"></x></message> > ``` yes it contains the word "discord" ↺

Gajim writes it usually to the log if something is wrong

I just experienced chat not scrolling down when posting while chat is scrolled up. I'm using Xorg. Just an FYI.

Not a complaint. I probably prefer that. It's just that sometimes it scrolls down and other times not.

is there a default timespan limit on chat searches in gajim?

there is no limit period

if you scroll you can search through all results

thanks

I was just a bit surprised at the speed, especially when selecting "search all chats", and being case-insensitive ✏

maybe it does a new search every time it loads more results

Gajim 1.4 to 1.5 was where we switched from "having widgets for _all_ chats rendered at once" to "render only what is displayed at the moment". This is way better regarding RAM usage, which significantly dropped from 1.4 to 1.5. Now since 1.5, to switch chats, gajim needs to 1) remove all widgets from the old chat (destroy) and 2) add all widgets to the new chat (db query + widget creation). Creating and removing widgets takes the biggest part. But there are certainly ways to optimize this. What I observed for the search view was for example that disabling the sorting function before removing all row widgets would speed things up a lot, because GTK would not try to sort after each removed row. I guess there is more, but the would need proper profiling and a bit of fiddling/testing :)

For me it's almost instantly

we dont disable sort

thats bad

:)

hm or not

we remove rows from the top, not randomly

why would a sort algo sort anything if you remove the top row

From the top means copying the full array minus one for each element removed?

Is it even using an array?

About this: I once used search to lookup an audio file, and it was far away in the chat history. I started playing it and then used the scroll-down arrow button, and audio playback immediately stopped

yes this is expected

how would you stop the audio afterwards?

go back to search and try to find it?

playback controls in the chat banner area :)

(joking)

yeah of course, lets add a dialog where you can add a music folder, and then you can play your music collection over gajim :D

Link Mauve, we are talking about Gtk.ListBox, i guess it uses a Gio.ListStore

> It provides insertions, deletions, and lookups in logarithmic time with a fast path for the common case of iterating the list linearly.

i doubt this has any noticeable impact

lovetox, g_list_store_remove_all() might be much faster than iterating and removing each element.

Although it could also just iterate internally, I don’t know.

g_list_store_splice() might be even faster, since it could avoid deallocating and reallocating widgets.

its not available on Gtk.ListBox

it does not expose the underlying model

Ah…

you can bind one yourself, but then you need to implement sorting and filtering yourself

we could simply destroy the whole listbox alltogether

and add a new instance

This might make a difference.

Hopefully this will use g_list_store_remove_all().

its not only about removing from a list

it needs to destroy all widgets anyway

> yes this is expected > > how would you stop the audio afterwards? > > go back to search and try to find it? the audio would play until it finishes :)

That would be _really_ bad without a button to stop it. Pragmatic way is to simply stop playback when jumping to the bottom.

Philipp Hörist pushed 1 commit to branch _refs/heads/master_ of _gajim_ < https://dev.gajim.org/gajim/gajim >: *03590b3b* < https://dev.gajim.org/gajim/gajim/-/commit/03590b3b75fad55d377135e4759cde96138d85aa > perf: View: Reset view faster

https://share.hoerist.com/philipp/qTteRxP0dkEaZYUs/Screencast_from_2023-10-02_23-13-31.webm

thats the speed of switching on my machine

seems fast enough

lovetox, we can’t see when you click to get to feel the latency.

On my laptop it clearly doesn’t feel instantaneous.

of course it isnt, and it never will

Why that?

because creating many widgets takes sometime

but, then its shiny, if you want instant use a console

and i didnt use my mouse, i used hotkeys to switch

otherwise i would be even slower

lovetox, then it won’t “never will”, it will start being fast once you don’t recreate hundreds of widgets but only the ones displayed, when using GTK 4.

lovetox, oh, using the mouse is slower?

i mean yes your input will be slower

if i hit a button 10 times, its faster then you moving your mouse from row to row and click

And yes i very much like the listview in gtk4, and sure it would bring performance gains

but i would say in the low 2 digit count

does this really justify rewriting 10th of tousand code lines

lovetox: great, that looks snappier

in my work i need to use MS Teams half of my day, its much slower then Gajim. But i never had the thought, oh i could get more work done if MS Teams was half a second faster on each action

i mean i see that big lag like 1+ seconds, is just annoying

but i dont feel it doesnt matter if you switch in 0.5, or 0.3 seconds

I don’t have user studies to back it up, but annoyance at a piece of software grows quickly with actions being slower.

I’m sure there are studies you could find about that.

And I’ve read a ton of people being extremely annoyed at MS Teams.