Gajim - 2023-07-15

hi guys when i chat with someone im gettin messages from strangers are rejected what do i do ✏

wakupp, does your server have spam protection

> hi guys when i chat with someone im gettin messages from strangers are rejected what do i do Add to contact and share online status with each other.

Mike Yellow, where can I join a fellow communist group I'm sick and fed up with capitalists and fascists?

wakeupp, explain more what u mean

why: did you solve your dns issue?

> Mike Yellow, where can I join a fellow communist group I'm sick and fed up with capitalists and fascists? There is no English communist public group chat on XMPP platform for now.

lissine, yeah it has been solved

> There is no English communist public group chat on XMPP platform for now. damn we should make one.

Hello. *Question*: Is there a way to tell the 1:1 chat window to remove the downloaded messages and again download the same from the server? (My Gajim version is 1.8.0 running in Manjaro Linux Xfce). *Background* I restored my Manjaro Linux VM to an earlier state (snapshot), and Gajim downloaded a bunch of messages from the server to a 1:1 chat window. Only, because I had a fingerprint issue, they were downloaded as error messages, "This message was encrypted with OMEMO and could not be decrypted." I have now fixed the fingerprint issue and would like to download actual message contents. Thanks.

Novice8: no

@lovetox, thanks. If I uninstall and reinstall gajim, will that work? (I presume on a reinstall I would be on blind trust if I can manage to uninstall all settings and config files, i.e. a clean uninstall.)

No you need to set back the encryption state

Which you can only do with a backup

@lovetox. Thank you, but I don't know what "backup" means in this context. But that's okay. It's not that critical for me to get back the old messages.

Instead, is there just a way to delete the error messages? I see "Delete message locally" for individual messages. But a way to do it en masse?

You can clear the chat history per contact

Check the contact settings

"Remove history" worked. Thanks. But did that remove history just from my 1:1 window or from the server as well?

Only local

There is no protocol to remove history from server

kind of surprising

XEP-0424: Message Retraction doesn't seem to mention if it can apply to only last message, single message, or multiple messages

but it seems to potentially remove messages from the server

For a future version of Gajim, it may be nice to be able to start the download (from server to your chat window) manually at a time you choose. Then you could enable OMEMO, take care of trusting fingerprints, etc. before the download happens. I understand the team cannot tweak the app for every contingency. It's a great app already.

Hi, what does gajim mean with "No devices found. Query in progress..."?

Novice8, the issue here is that OMEMO is a protocol with forward secrecy, which means the keys used for encryption change for almost every message, and your client throws away the old key when it receives the next message, making it impossible to decrypt the same message twice.

You seem to want an encryption mechanism without forward secrecy, so check out OX for instance.

It uses the same key for all chats, an obvious downside is that if an attacker manages to obtain your key they can decrypt all past, present and future messages.

So it’s up to you, whether you want to weaken your encryption and allow rollbacks in time and new devices to be able to decrypt past messages, or if you want forward secrecy and thus pick OMEMO and cry for lost history whenever something bad happens.

@Link Mauve. Thanks for the heads-up. But on "impossible to decrypt the same message twice," that cannot entirely true. When restored my Manjaro Linux to a date in the past, Gajim downloaded and decrypted the messages in all the 1:1 chatroom that did not have a fingerprint issue. That's the second time, the first time being when the messages went back and forth real time. The decrypt failed only in the 1:1 room that had a fingerprint issue (I had to re-trust it). So, a way to tell Gajim, "Hold your horses till I take care of the fingerprint," would make it possible to download and decript past messages a *second* time even if there is a fingerprint issue.

By "decrypting the same mesage twice," you must have in mind something else, and I am sure I was not suggesting that that be changed.

Novice8, there is no room involved with 1:1 chats, fyi.

It all goes through directly to your contact.

Sorry. I should have said "1:1 window."

But no, I stand by it, when your client changed its OMEMO key, you can’t roll back.

"changed its OMEMO key." Ah. That's not my scenario. When I roll back the machine to a time in the past, the keys are all the same. Only, there was a problem with a contact's fingerprint.

Since you seem interested, I will tell you the whole story. The problem was that, even though the contact's fingerprint had been trusted, the decrypt failed. This happened let us say on July 1. I fixed (per instructions I got in this room), on July 5 (it basically involves deleting the contact's fingerprint and sending a message and re-trsuting).

But on July 15, I turn the clock back and restore my Manjaro to the state it on June 1. Now that before July 1, but it so happened that the same problem was there again.

So, when I clicked into the affected 1:1 chat window, the usual download from server happened, and the decrypt failed, obliging me to use the same fix again.

So for the whole time, there was no change of OMEMO key. And if only the 1:1 window said, when there was history to download, "Hey, you've got history. You want to download now?" and you could say no and fix the problem before downloading, the affected 1:1 window would behave just like all the other 1:1 windows. I.e. be able to download and decrypt history just fine using the same respective keys.

Novice8, you still seem to believe OMEMO keys are immutable, they are not.

Now, that's all I meant. And since this redownloading and redecrypting of history happens all the time, I am sure it's not what you meant when you said something was impossible.

On first exchange with a new contact they will pick one of the many prekeys you published, and on every subsequent message you will throw away the key and derive a new one.

There is no redecryption possible with OMEMO.

If OMEMO keys change on their own all the time, that's fine with me. All I meant by "same" was that I didn't delete or change any key.

You don’t do it manually, your client does it for you on almost every message.

Fine. I will say that redownloading and the representation of legible messages happen all the time and cannot be impossible.

It just happened today before my very eyes. The same OMEMO encrypted messages are presented to me for a second time in a legible form. And I could cause that to happen a 100 times just by restoring the computer to a timein the past. You are going to deny the term "decrypted" to an encrypted message that has become legible to human eyes?

Ok, I’m out of this conversation then.

I am sure you had a valid technical point which, when properly stated, is consistent with what happens all the time before my very eyes.

You can decrypt old messages if you set your omemo state also back in time

Which is only possible with a backup of said state

If you use a old backup of your VM I think decryption should work, would be interesting what the actual error in the logs was

But fingerprints don't matter for decryption

They only matter when we decide to encrypt for what device

I'll be happy to upload my log if you tell me where it is.

(Manjaro. Gajim 1.8.0,)

Link Mauve: the omemo messages in MAM are static, they were already encrypted, with some session key(s). if you rewind your session key on the receiving side, you should at least be able to decrypt the messages that were encrypted for that session key. I'm not really sure if session key rotation is deterministic, so dunno whether it would rotate into a different state in the new timeline. if it is deterministic, maybe it's possible to decrypt all of the following history

but I'm inclined to think it's random

so the messages that happened to be decrypted "twice" were all encrypted with one session key

mjk to my knowledge there is a ratchet

and this is changed each time i send a messge to the contact

hm i asked some other colleagues, and they told me that im wrong, backup/restore does actually not work

or only for the very first message

so to simplify it, if you only receive messages, this could work with a backup, to receive and decrypt them again

but as soon as you reply, the session state is non-deterministicly changed

and you cannot replay this

so it can look like some messages are decryptable after backup/restore

so in general we can say, except for a very limited edge case, omemo messages cannot be decrypted twice, even with backup/restore

In that case (with so many voices against my case), I must have been wrong. I will let Gajim accummulate messages for a while and try another restoration. If I am wrong, the feature I requested above cannot exist.

> but as soon as you reply, the session state is non-deterministicly changed thanks for confirming 👍

Novice8, i dont think you need to test this

also message your wrote yourself are impossible to decryp

because they are not encrypted to "self" in the first place

^ good point

only the recipient devices can decrypt them

(including your own devices, but excluding the sending device)

what i notice Gajim becomes conisderable slower if my laptop goes into Power Saver mode

probably normal, but shows that some actions like switching Chat are CPU bound

I think that was always reported as a computational issue ^^'

but to add a +1 to it, switching chats maxes out one of the raspberry pi 4's cores here for over a second

recently I actually wanted to suggest keeping a cache of 2 or 3 populated chat views, so that at least switching between often used chats is fast

I also wonder if keeping a pool of constructed message widgets and just changing their content would improve perf when populating chat view

a reminder that constructing a new code snippet widget takes about 50 ms here

other widgets are much less intensive, but are still in milliseconds

this is insane, but idk if it's fixable from Gajim's side

mjk, fyi that’s exactly what GTK 4 brings to the table.

And part of why it’s so much faster with extremely long list of widgets where only a small fraction of them are displayed.

cool, but... I think Gajim only loads, what, 20 messages at first? if *that* takes more than a second, I don't think any list widget magic would help

We load anyway not much that's not on screen

yeah

What makes it slow then?

Slow is relative, it depends on the CPU power and when we talk about it we never talk about the CPUs used

lovetox, mjk mentioned it was on a Cortex-A72.

ok and how do i compare this now to what i have?

You can’t, barring buying the exact same CPU.

But most CPU-bound tasks scale approximately the same no matter the CPU used.

That is, if you improve the task by 50% on your CPU, it will improve it by approximately 50% on mjk’s CPU as well.

ok .. i would have expected that

but this does not help

if i use a cpu benchmark site

Those are mostly garbage tbh.

a cortex A72 2GHZ has 720 points

my cpu

Intel Core i5-8250U @ 1.60GHz

has 6000 points

so its more than a factor of 10

and my laptop is 4 years old, and was never a high end model even back then

lovetox, yup, sounds about right.

Cortex-A72 is an embedded CPU from 2015.

ok, good, see thats what i mean, we need to start trying to get objective data for these reports that something is "slow"

and sometimes the answer is, yes you have a slow ass cpu

lovetox, well, sure, but I’ve seen people with much better CPUs already complain about Gajim’s speed.

So if you improve it for everyone, even such slow ass CPUs will benefit.

of course, i will try to improve once i see that something is inefficient

just to add a couple more cents: chat switching is slow _relative_ to pretty much everything Gajim does (not counting python compilation at startup). e.g., scrolling the chat view is very smooth and doesn't have any noticeable lag. and that is being rendered on cpu

resizing the window horizontally could be smoother, but it's still multiple updates per second

mjk, Python compilation is usually pretty fast, and when installed it doesn’t happen at all since the .pyc files are already on disk. From previous benchmarking I did, it is loading the module files at all which is slow.

Link Mauve: loading the modules is more than compiling?

Yup.

ok

And it only happens once per module when it gets modified, which is a pretty rare event.

On subsequent loads the compilation won’t happen.

once per module per process lifetime?

Not per process lifetime, per the .py file hasn’t changed on disk.

ah

then I have no idea why literally every python program I run has time-to-any-signs-of-activity in multi-second range

of course hard disk could also a problem, if you run debug you will see how much the database query consumes

but db isn't cpu bound in Gajim, is it?

its not, but if you say normal loading of messages while scrolling is fast

maybe the problem is something else

as for python startup, its slow, the reason is that python loads many modules

no, not loading, I mean scrolling the view with already loaded messages

but if you scroll, more messages are loaded

loading more messages from the db is still noticeable, but not a whole second

why would scrolling be slow? its not the gtk drawing that is slow

its the object creation the widget creation that is slow

> why would scrolling be slow? its not the gtk drawing that is slow because of the "slow-ass cpu" :P

i think drawing stuff in a scrolled view is trivial

I'm pointing out that it's plenty fast for all activities except chat widget creation

even on a old cpu

🤷

its the widget creation that can be slow, because we run much code until a widget is finally created

and also the gtk layout engine, if the widget is shown, it needs to be calculated with all other widgets etc etc

but in a scrolled view, simply scrolling, there is no resizing going on, there is no creation going on

its just the info what pixels need to be drawn next on some canvas

ok, but as I said, resizing the chat horizontally is still pretty fast

> gajim.c.storage.archive Execution time for get_conversation_before_after: 28 ms that's the greatest I see

looks not bad

and often it's less

and you say it takes one second to switch?

> its the widget creation that can be slow, because we run much code until a widget is finally created ohh, right, there's a bunch of python code

> and you say it takes one second to switch? yes

lovetox, note that on a Cortex-A53, which is even slower than mjk’s Cortex-A72, scrolling with GTK 3 was slow and I managed to make it about twice as fast in GTK 4 using its GL renderer.

Despite my GPU also being quite slow.

It was about 20 fps on GTK 3, at 720×1440.

Link Mauve, im sure GTK 4 has improvements

yeah, I also remember scrolling on a raspberry pi 3 being unbearably slow

It was just an answer to “20:00:58 lovetox> why would scrolling be slow? its not the gtk drawing that is slow”.

Is there any work being done or plans on migrating to gtk4?

No.

no

Just not enough resources for now or is there something wrong with it?

cost / benefit is not right

its better to implement features that people actually use for chatting

working a year on a gtk4 port, to tell people, hey we worked a year, and some of you on slow cpus might feel its now a bit faster

is not worth it in my book

there are even dealbreakers like no status icon anymore on Windows

"oh btw, you don't have a spellchecker anymore" :D

and i bet X more if i get into the docs what they all removed

probably a handful features would need to be removed from Gajim to make it even work

yeah, I vote for features and slow-but-steady perf improvement

to that effect, reactions when? :D

yeah .. i need to finish the database rework

right. godspeed!

but it will be a big big change, and i hope the migration will work

uh-oh, does Gajim need integration testing now?

we added some database tests now

but someusers have really old databases

and i think it will be a suprise what we find in some :D

lol

maybe require an upgrade to 1.8 before upgdading to 1.9?

random thought regarding performance: I wonder how much time _cleaning up_ a chat view takes. probably less than creation, but I've seen some unexpected things in my life

mjk, you can profile. :)

yes thats also a place i would look into

because switching means cleaning up first

> mjk, you can profile. :) nahh, too busy procrastinating

Why does my contacts avatar image suddenly get missing in my Gajim? But not in my Dino? I'm pretty sure, my contact did not change anything in their setup (Conversations, jabber.de), because they is not very "technical". ✏

mjk: right, clean up can be expensive as well. For the search view for example, I noticed that disabling the sorting mechanism before removing rows would speed up the clean up significantly.. But yes, profiling helps pinning it down. The wiki has some answers on how to do that. It's actually quite fun to view it in Snakeviz :) and rewarding if you can improve speed!

sometimes I get "Error: You are not currently connected to this chat" in the chat view

1. what causes this 2. why doesn't the disconnected status icon show in the chat list 3. if gajim knows it's disconnected, why doesn't it reconnect?