Gajim - 2025-04-18

curious, will Gajim have any plans for parsing xmpp: URIs especially for account invites?

> curious, will Gajim have any plans for parsing xmpp: URIs especially for account invites? 👍 ↺

https://dev.gajim.org/gajim/gajim/-/issues/11745

Ok so the plan is there at least 😅 ✏

Does Dino support this?

Hmm apparently not

No https://github.com/dino/dino/issues/232

Ugh, very stale issue 😢

> https://dev.gajim.org/gajim/gajim/-/issues/11745 Huh, there has been no further activity whatsoever? But anyone can still contribute to that issue? :o ↺

Well, if there was no plan to add that someone would have probably closed the issue by now. ✏

who closes issues ? gajim has 240 open and dino has 530 :)

How do I select several - but not all - messages in one conversation and delete them all at once?

moparisthebest: hm and over 10.000 closed?

Good morning folks, I saw that there is a feature block contact ("Kontakt sperren"). I was wondering, is there also a way to unblock contacts?

egnun: check the accounts advanced menu

dln9mobster: Does message menu select messages then delete not work? ✏

>> I was asking because of how fast you translated it 😁 > ah, no i'm just a strange person and I read lot of XMPP dev channels :D ❤ ↺

> moparisthebest: hm and over 10.000 closed? 👍 ↺

Hello all

is it known that gajim re-orders messages sometimes?

poezio shows me a different order for messages in a MUC than gajim.

this has caused quite the confusion right now.

hello, I've noticed that after a while that gajim is running in the background, the tray icon suddendly disappears even though the application is still running

and you have to close and reopen to make it appear again

is it a known issue?

jonas’, depends but yeah this can happen in multiple scenarios

if you compare side by side clients

maximus, sounds like GTK4

emru, you think it's a gtk issue?

no GTK4 does not provide a tray icon anymore

oh ok

so its either a issue with the app that draws the draw icon

but if you had for a moment, and then it disappears, I think it's something different

or some problem with the dbus connection in gajim which should show in the logs

I use waybar btw

lovetox, okay, why? MUC has a fixed message order, why does gajim reorder messages?

jonas’, one reason I could think of is 0203 delays, another is 0308.

Link Mauve, neither was the case (delays are shown by poezio differently, right?) and the message wasn't edited.

Indeed.

jonas’, we sort by timestamp

with which precision?

its a integer

that doesn't answer the question :D

depending on the server

what does the server have to do with that?

because it decides the precision

where does the timestamp come from, then?

i can come from the server or the users machine

normal message stanzas don't have timestamps

and when it comes from the users machine, which precision do you use?

whatever the machine gives us, usually ms

so those messages must've arrived in the same millisecond to be reordered?

thats impossible

because message processing takes more then 1 ms

does gajim store the original, unaltered message XML somewhere?

oh

wait

I got those from MAM

right

so the MAM likely has second precision

and gajim does nothing to disambiguate

are there plans to fix that?

why are you assuming that

assuming what?

what you just said 2 sentences ago

that gajim does nothing to disambiguate?

yes

well because the order I see in poezio differs from the order I see in gajim.

(and the order other people see also differs from what I see in gajim)

yes still assumptions

alright, what *does* gajim do to disambiguate?

it sorts by time and insertion order in the database

but your theory is easy to test, create a muc on the server, send 2 messages in the same second (should be easy) watch MAM stanzas flow in

which column contains the insertion order? I only see `timestamp`.

(in the `message` table)

i just got pinged, lmao

sorry for that :D

the column "pk"

that's the rowid

in sqlite

without AUTOINCREMENT, sqlite may reuse rowids

(that the pk got mapped to the rowid can be confirmed by looking at the output of `select rowid, pk from message where rowid != pk limit 10;` and `select rowid, pk from message limit 10;` (where it shows `pk` as column header for both). and the claim that without AUTOINCREMENT, things may get reused is noted here https://sqlite.org/autoinc.html )

so apologies for my assumption there, I didn't expect to fall into the rowid rabbithole today.

gotta run for lunch now, see you later.

ok before i read into that, it would be interesting what this query returns on your machine

hmm, Gajim 2.1.0 (Arch Linux) doesn't want to remove account or a server that doesnt exist anymore, and Disabling the Account first doesnt work; it switches it back on when my back is turned. Removed an account that does exist just fine. Just a heads up, think I'll nuke it and pave over

nigel, bevor you pave over, please check the debug logs and send me what it says once you try to disable the account

can do.

disabling an account should have nothing to do with the fact if this account is sable to connect or not

umm, never done that before haha

seems like pasting files to the gajim window with ctrl+v is not working anymore

preferences -> advanced -> enable debug logs, and restart

(gajim:56192): Gtk-WARNING **: 11:28:54.082: Trying to measure GtkGrid 0x562fa8769f50 for height of 260, but it needs at least 1339 Traceback (most recent call last): File "/home/gajim/gajim/.venv/lib/python3.13/site-packages/gajim/gtk/message_actions_box.py", line 884, in _on_clipboard_read_value_finished file_list = clipboard.read_value_finish(result) gi.repository.GLib.GError: g-dbus-error-quark: GDBus.Error:org.freedesktop.DBus.Error.AccessDenied: Invalid transfer (9)

oh ffs.... as soonas debugging turnde on, it worked instantly. ✏

oh ffs.... as soon as I turned debugging on, it worked instantly. ✏

I had restarted client several times and it still would not disable nor remove.

debugging on, restarted, account, Remove, succesful instantly.

mrdoctorwho, no idea what that error means but on first glance i would say thats not a problem in Gajim

lovetox, the same file can be pasted to chrome, but not in gajim for some reason

I also upgraded my distro release, maybe that could be the cause though

> preferences -> advanced -> enable debug logs, and restart DM sent with log file, in case it is useful. Thanks. ↺

Weird that it would use DBus for transfers, I would expect Wayland to be used instead.

mrdoctorwho, are you using flatpak?

We request a file type Gdk.FileList

docs says, Gdk will try to convert content if it does not fit the type, i assume something fails there

lovetox, no, but gajim is running under a different user, plus using xwayland, which likely does not have access anywhere and dbus is not running there. Although the file I'm pasting has the 777 access. I recall it was working just yesterday, but today I upgraded both fedora release and gajim

we didnt change anything on the code for a long time

access denied seems to hint to a problem with permissions

lovetox, could you maybe suggest where to look? As I said, the file has 777 permissions, plus the same file can be pasted into chrome without any issues

chrome is running under a different user as well

so there is the user A which has copied the file, then there are users G and C (running gajim & chrome). Gajim can't retrieve the file, but Chrome can

no sorry, i have no clue about linux internals regarding how clipboards operate

all we do is call a API from GLib Clipboard.read_async()

and the error tells us that we are not allowed to read the clipboard for whatever reason

I'll add a fix to catch the error (same for pasted text)

cal0pteryx, in case of exception please also print a debug log what the content of mime_types is

if you copy the file to clipboard under the gajim user, it works fine

lovetox, https://paste.debian.net/hidden/1a8d7598/

i dont understand the query, it says that row_id = pk if we dont use autoincrement

so what sense would it make to select row_id != pk?

it was meant to prove exactly that ✏

it return no results, and thats what we would expect here

I wanted to make sure that the table indeed uses rowid for pk

before making the claim that this might be a problem without AUTOINCREMENT

(having been burned by unfounded claims before today :-)) ✏

i read a bit into it, and you are right rowids can be reused, but the way they would be reused is not a problem for our sort algo

the standard row id picking is still incremental from the max row id in the current table

and thats whats important

its just that when you remove the last 100 rows, it reuses these 100 row ids, but again still incremental

with the only exception from that rule if you go over MAX ROW ID

which is a number so big that you would need to pull all messages from all xmpp servers in the history of humanity

that seems correct

makes me scratch my head as to what went wrong here

the rowid of the first message is larger than the rowid of the second message.

which would tell us that we received the first message later, and its indeed not the first message

is there some kind of gajim XML console?

yes, under View menu

ah, the protocol tab in the debug console, thanks

and do the 2 messages have the same timestamp in the database?

yes

so

the MAM response indeed has the wrong order

I rest my case and stand confused.

did you type the messages and send it, maybe its time to question that information

I don't understand that question

did you send the messages yourself? as you are so sure about the correct order

no

but three other clients agree on the order

it's just my gajim from mam which disagrees

so the other clients might received the message live

which is still weird, that the server would insert it with a different order into the database

but as said should be easy to reproduce, sending messages in the same second can be done by a human

jonas’: my Gajim, Dino and Conversations show the same order of the confusing two messages fwiw. First 2.jpg and then 1.jpg All fetched from mam

lovetox, if that was indeed the reason and not some complicating circumstance. in any case, gajim is doing nothing wrong here. Thanks for the patience.

Hello everyone. I just joined this group because I have a question about Gajim. If I upload a file on an XMPP server in a private chat encrypted with OMEMO, is the file name encrypted (or changed) as well?

jonas', FYI prosody 13 supports sub-second precision timestamps

wildegim, no, filename stays the same

and is not encrypted

there are better long term solutions

> wildegim, no, filename stays the same Ok thanks. Maybe an option to at least randomize the name would be a good idea? 🤔 Has it ever been explored as far as you know? ✏

> there are better long term solutions Which ones? 🤔

randomizing the filename will be part of the long term solution, but right now we cannot send the original filename with the message, which means its lost

someday we will be able to send that, and then you can have both things, random filenames on the server, but real filename in your client

> someday we will be able to send that, and then you can have both things, random filenames on the server, but real filename in your client This is much better than randomization only... Does this require a support from the XMPP standard or only from the client?

> someday we will be able to send that, and then you can have both things, random filenames on the server, but real filename in your client Will this be a new spec (based on XEP-0454 or entirely new) or is it already specced?

It seems like that the relevant pull request has been closed: https://dev.gajim.org/gajim/gajim/-/merge_requests/588

Hi, has anybody experience with EXWM here? I am experiencing issues after updating to gajim 2.0 (and later) regarding the send-key functionality of EXWM. It looks to me like the user interface toolkit GTK 4 and EXWM are not interacting well. The issue is that I cannot send control keys, e. g. CTRL-C, to gajim anymore. It worked well before gajim 2.0 and it still works with other applications like firefox. Any hints highly welcome! Thanks.

wildegim: randomizing the filename will keep it a secret from the server, but will also keep it a secret from the receiver. A better thing would be to randomize the file name, and send the original filename omemo-encrypted in the upload's metadata

Though I'm not sure if this has been proposed as a protocol extension yet.

Randomizing the filename can be a good intermediary step for now.

Talking about this: https://github.com/emacs-exwm/exwm/wiki#simulation-keys

> wildegim: randomizing the filename will keep it a secret from the server, but will also keep it a secret from the receiver. > A better thing would be to randomize the file name, and send the original filename omemo-encrypted in the upload's metadata Yes, I have understood this. I noticed now that the pull request I linked does only half of the job. I would propose to open a full issue on the Gitlab to discuss on this, because I believe this feature is of utmost importance, since I notice that uploading a file on a server, exposes it via HTTP to *everyone*.

Which normally is not a problem if it is encrypted and the encryption key is not present, but if the filename reveals its content (e.g. a popular image on the internet, "political-party-A-manifesto.pdf" and so on) encryption becomes useless.

wildegim, everyone who knows the URL, which should be only you and the recipient(s) and the server admin who can do a ls in the right directory.

That’s far from everyone.

EXWM-user, as i understand it this happens before the toolkit gets it, as it translates some keys before sending it to the window

> egnun: check the accounts advanced menu lovetox: Thanks. The list ist empty. Also Dino didn't show the contact as blocked.

egnun, then the contact is not blocked.

EXWM-user, maybe check other GTK4 applications, to see if it affects other apps as well

Ah, but if I try to block a contact again, it now gives me the warning again.

and after you block it, and reopen the dialog, the address does not show up in the blocking list?

lovetox: Thanks. Yes, it uses xcb:SendEvent to simulate a key. The thing is, that this worked with gajim < 2.0 and stopped working with gajim >= 2.0. And the suspicious change is about using GTK 4.

lovetox: Good idea. I will try that.

> That’s far from everyone. But by knowing the URL everyone can check. If the server has a security vulnerability that allows for directory enumeration users' privacy is at stake. For instance, conversations.im saves all files sent by a user under a directory called <username> and the link has the structure share.conversations.im/<username>/randomstring/<filename>. If I, as an attacker, can somehow do an ls -R I can see every file uploaded by <username> with cleartext filenames. ✏

wildegim, yes using a random file name is better security

but using the real one is better user experience right now

Maybe use a random filename for encrypted uploads, and the original filename for unencrypted ones.

lovetox, since the final idea is to have both (i.e. the server seeing a random name and the user the right one), would it be a good idea to open an issue to track efforts in implementing this?

https://dev.gajim.org/gajim/gajim/-/issues/11470

there is already an issue

I get a Gtk-WARNING when hovering the edited icon ``` > Failed to set text 'Message corrected. Original message: > That’s far from everyone. But by knowing the URL everyone can check. If the server has a security vulnerability that allows for directory enumeration users are screwed. For instance, conversations.im saves all files sent by a user under a directory called <username> and the link has the structure share.conversations.im/<username>/randomstring/<filename>. If I, as an attacker, can somehow do an ls -R I can see every file uploaded by <username> with cleartext filenames.' from markup due to error parsing markup: Error on line 5 char 46: Element “markup” was closed, but the currently open element is “username” ```

> https://dev.gajim.org/gajim/gajim/-/issues/11470 Wow, thanks 😊

lovetox pushed 1 commits to branch gajim/master fix: Chat: Don’t fail in some cases to show correction/error tooltip - https://dev.gajim.org/gajim/gajim/-/commit/d0ae74a0b2f77c07f9a426546fb7fd9df547e5de

fixed hannibal

thanks

question regarding creation of group chats in Gajim: is there a way where you have a group chat that has this: 1) people, who aren't part of the server in which the group chat is based in, wishing to join inside need to be invited by a member of said server 2) the group chat is however visible for anyone that is a member of said server and even is visible to them for them to join?

1) if your server federates with other servers it does not matter where a groupchat is, and yes you can configure a groupchat to be invite only

luna: that's solved on the server side

2) there is a global groupchat search in Gajim, some clients offer you to search your home server for groupchats, but they still need to be invited

ohhh, so like, I can't have more finer grained control regarding who gets in?

no, just invite only or public

dang, but yeah

luna: you could make the group invite-only, add everyone in your server to the group, and allow them to invite other users

But once users from other servers join, you can't prevent them from inviting other users

oh huh

you could write a prosody module that allows you to restrict the group invites permission to the group's admins. Then you can make all your server's users admins in the group, and you'll have the desired behavior ✏

but isn't that kinda a hack?

have there been XMPP proposals to expand more upon permissions regarding chat?

your "permission only for people on my server" is not a use case i ever heard

people invite other people into groupchats and manage them on their own

they should not need to care on which server another user is registered

I mean, I'm big into having much more customization but fair

because I'm mostly worried about moderation if I have a public chat that is listed

in that area, are there even like, moderation bots for that, or similar?

what does a moderation bot do?

filter out any messages containing inappropriate words? spammers?

servers have anti spam measures, like limits on how much you can send

but yeah, if you operate a big chat, you need people to moderate it

like any community

i mean, isn't that exactly also good, having a bot on top of moderators?

nobody stops you from writing a bot

okay... but are there any, if someone has already implemented them?

maybe, im not aware of any

i think you overthinking stuff a bit, think about that if you manage to get a groupchat that has so many users that something needs to be automated

this is a groupchat with 400 people, and we never had the need for a moderation bot

ah fair, i prolly don't need it if i don't have many people in the gc

as long as they are chill so yeah!!

you are in luck, xmpp is not that popular that you need to fear that many people join your chat :d

oh dam, yeah!!

but yup, I usually always think first with security and privacy in mind before anything else

as you can see that c:

thank you for letting me know in case!!

RE rejoining room issue: I enabled `mod_private` in Ejabberd & everything seems to work now. Very strange.

toastal, i fixed a bug recently that can lead to not joining rooms after standby

Neat

Also neat: hello, world from Cheogram from Sailfish OS’s AndroidApp Support

I recently noticed that typing a few letters of a user who is in the chat and hitting tab no longer auto completes their usernames (instead I get a tab character). Any suggestions on how to debug this problem?

jamie: its removed in 2.0

jamie, try "@"

Now you have to @ like the bad systems, slack etc

ahhhh. I see - thank you!

> dln9mobster: Does message menu select messages then delete not work? It does, but no option to delete appears, and pressing the deletion key also does nothing. Deletion option appears only when I select only one message. ↺

in the message menu, it says "Delete messages locally"

ah yeah

hm seems we didnt implement multiple delete yet

do you really want to delete specific messages, or everything?

we have a option for everything

lovetox: You are right. It looks like my issue is a thing between exwm and gtk4 in general. Because I could reproduce it with another gtk4 application and not just gajim. I will file a bug for exwm. Thanks a lot!

> do you really want to delete specific messages, or everything? Well, sometimes all messages (but without deleting the conversation), and sometimes a number of messages, but not all. ↺

all messages, you can do in the chat settings

multiple message delete is on the feature list

> you are in luck, xmpp is not that popular that you need to fear that many people join your chat :d ❤ ↺

> Also neat: hello, world from Cheogram from Sailfish OS’s AndroidApp Support 👋 ↺

lovetox pushed 1 commits to branch python-nbxmpp/master change: Retraction: Rewrite parsing code - https://dev.gajim.org/gajim/python-nbxmpp/-/commit/b9d6b25c55ec437294a998291da3018f7eca8957

lovetox pushed 1 commits to branch gajim/master cq: Message: Add type annotations - https://dev.gajim.org/gajim/gajim/-/commit/1550705a99c23717fb736e6d3103a4dde0a2e036