Gajim - 2024-06-05

I joined your test MUC but reply button isn't displayed there somehow, hehe. Why? Does it rely on some feature not advertised there, like persistent IDs or so?

How do I properly transform account string into client string? https://conference.gajim.org:5281/pastebin/1221139d-96cd-43b7-91d7-3028b8f86e89

I wonder how did it even determine by RDNS that it's tif.bot.nu.

(I take account strings from `app.settings.get_accounts()`)

Hehe, probably need to replace with `get_clients()` obviously.

> I joined your test MUC but reply button isn't displayed there somehow, hehe. Why? Does it rely on some feature not advertised there, like persistent IDs or so? ``` if self_contact.role.is_visitor: return False ``` Interesting. Why? Safety? ↺

> Attempting to retract a message [using the overflow menu - Retract message button, in a room where one is an admin] in Gajim is causing a crash, but one where I can't even copy the traceback nor click on Submit bug report...after the crash, Gajim just freezes and has to be killed and restarted. v1.8.4 on Debian Testing. Ah, so the focus was on the "reason for retraction" dialogue (which was hidden by the traceback window). That's why I couldn't access the traceback text, nor click on the Report Bug button.

the server won't allow visitors to reply

## Versions: https://conference.gajim.org:5281/pastebin/d99d618f-5939-4995-8296-77f9456706f1

lovetox: traceback ☝️

concerto: are you running 1.8.4 or master?

cal0pteryx: The Help - About dialogue says v1.8.4. The Debian package is gajim-nightly 20240410-1 ✏

concerto: it looks like a version mismatch between gajim and nbxmpp. Make sure both are up to date

> this effective power txt meme still crashes gajim tfw lol, haven't heard of that bug in nearly 10 years ✏

funnily enough we were just bringing it up recently after finding some discord crash bug we found

and the bridge crashed my gajim

hmm, I can't figure out where the voice recordings temp files are kept. lsof says they are in /tmp but it's not actually there.

They are deleted

It's weird because I'm watching with `inotifywait --monitor --recursive /tmp`, then start recording, then do `lsof -p 1234,1235 | grep m4a`. lsof shows that a new file exists in a gajim directory in tmp, but it doesn't, and inotify wait has not yet seen any changes. ✏

How do I make the following autoplay in a loop, without sound, in my friend's group chat ?

https://chat.domn.net:5443/upload/500b07a7efeff7dd5a382d5bf3fa0d30cec686c6/PBIlJyG36gKzdoJVcxog35rqziUTP7HO9RpYx32t/firefox_oTDBwGTPiX.mp4

Without them having to click on it

And it then opening in an external program ? :\

You don't

gajim as full blown media player when?

Surely GTK can display 3d models, video streams as easily as text. I'll ask chatgpt

Ah, there we go https://chatgpt.com/share/f5564e03-5f0d-4a7b-a265-b806779d3bae

Oh, it uses ffmpeg to convert the mp4 to gif first, gross ! Surely, there is a more elegant way

I wonder if I'm an outlier that I wish the voice recording filename had time in UTC to preserve location privacy

gstreamer does a thorough job of adding metadata :)

> Oh, it uses ffmpeg to convert the mp4 to gif first, gross ! > Surely, there is a more elegant way Lottie renderer would definitely be useful. ↺

fjklp: I believed you generally send voice messages to your friends and not strangers, so the privacy concern is not really given? :)

fjklp: it creates files in `/tmp/gajim-{4 char random string}`

but as said above, the folder is cleaned up when Gajim exits.

``` https://conference.gajim.org:5281/pastebin/6260ad01-8756-4717-8eaf-d1ac8bd7fb7c

> fjklp: I believed you generally send voice messages to your friends and not strangers, so the privacy concern is not really given? :) that's what lsof tells me but it doesn't exist ↺

ffs

replied to wrong message

bodqhrohro: look at the logs it will print warnings if for some reason the stanza I'd is not accepted

fjklp: you could have corrected that

fjklp: but it does send a playable audio file?

You can run gajim.with debug log enabled, it will tell the path when it creates new files: `gajim -l gajim.gtk.voice_message_recorder=DEBUG` IIRC

is there some manteiner for the aur package of gajim-git?

for some reason there aren't further updates of the package since sometime in may

> fjklp: but it does send a playable audio file? yes, it seems to be working normally ↺

fjklp: so what's the issue you are tring to solve? 😄

mostly curiosity

multiple questions occur to me

is it written to disk? uses the systemd tempfs? in memory? when does file cleanup get done, if ever?

I've come to expect file cleanup should not be assumed

I've mentioned it here already a few days ago, anyway: Files are written to disk into a temporary folder. When you close Gajim, these temporary files are cleaned up as unlikely that you'd still need them since you should have sent the audio files. This design was chosen as it was the easiest to work together with the audio player widget. It should also be possible to do it all in memory but that requires some more work/time and knowledge I'd first need to teach myself. :)

It'd should be easy to use memory files using memfd_create but that's unfortunately a Linux only solution.

> I've come to expect file cleanup should not be assumed Why?

> > I've come to expect file cleanup should not be assumed > Why? Various experiences of finding files left that I felt probably shouldn't. One I can recall was cargo having multiple gigabytes of cached files from years ago. I just found a fairly large file left in ~/.cache/obexd from what I assume was a partial bluetooth transfer years ago. ↺

Regardless, I tend to want to verify rather than assume

Bad implementations or crashes happen, prohibiting a proper clean up in the end.

hmm, I think I'm experiencing a bug here, I can't autocomplete meson for some reason

but I have another account in here that *will* autocomplete his nick

maximus: I think Polarian managed that?

maybe multiple accounts causes a problem with this?

managed what?

cal0pteryx: I think its Link Mauve

not me... I was the tester for `gajim`

well, it will but it's doing it out of order. meson/2 , then meson

sorry for all the mentions

I'm still joined twice? Woops

No, Im not..

meson, I wanted to say thanks for the contribution. I think it's a nice addition to gajim :)

fjklp: there should only be "meson" around, "meson/2" was only here a few days ago for a limited time.

fjklp: great you like it! :) ✏

isn't tab completing nicks supposed to cycle through nicks in order of most recent post first?

fjklp: yes, that bit me, too. Must have a look at the logic ✏

I have a note in my gajim bugs file since way back to deal with that but I haven't experienced this exact thing that I recall

I'm pretty sure having multiple accounts in the same chat has an effect. Each account will behave differently with tab completion.

I really like to figure out what's up with gajim and image previews

when automatic image preview is disabled, isn't the preview box supposed to show an image icon that you can click on to show the image? For me, it's showing a question mark in a box. Clicking that icon does make the image show up.

fjklp: check what `xdg-mime query default image/jpeg` outputs maybe?

How come that Gajim uses the character "," when mentioning someone? The most common (and probably the only character I've ever seen) is ":". :)

pep., I think you were interested in an "ignore autojoin setting", it seems to work for me but if you want to test it it's always nice ;) <https://dev.gajim.org/gajim/gajim/-/merge_requests/1053>

Also. Why is the registration form on https://translate.gajim.org/accounts/register/ partly in Chinese? :P

https://share.snikket.linuxkompis.se/upload/YGLK3doyTsnpHVf24ako2ViG/1573bf2d-2669-4cd2-a0f2-034442b8577e.png

> How come that Gajim uses the character "," when mentioning someone? The most common (and probably the only character I've ever seen) is ":". :) hexchat uses a comma by default. I think colons make more sense. ↺

fjklp, Okey. :) I think it would be nice if the user could choose for themselves.

Hund, you can in the advanced configuration editor

Look! That's pure horror. It looks so wrong! :D

Ohhhh

fjklp: Wow. Thank you!

Hund: you are welcome

https://upload.gajim.org:5281/upload/VBLHLW5exzdgArhQi-SkF2FX/8b8582f2-247b-41c9-99bf-4b9a1dd32438.png

this is a screenshot of how I see the last posted image in this chat

xdg-mime: image/png mimetype: image/png

fjklp: You need to enable image previews in public chats.

for public chats*

what I'm expecting is that the question mark icon would instead be a generic image placeholder icon

Oh

it might be even better if we just had a button that plainly says "View Image"

That would be a lot better.

I would love to see quotes to have some visual tweaks as well. :)

Hund: tweaks?

> it looks like a version mismatch between gajim and nbxmpp. Make sure both are up to date cal0pteryx: I seem to have gajim-nightly 20240410-1 (is this really the most recent nightly version?) and python3-nbxmpp-nightly 20240605-1 ...but the error still persists 🤔️

Another trace - https://conference.gajim.org:5281/pastebin/ba7dccd2-abc9-4bdf-956f-a39e82ab5607

About previews: I wanted to look into also deleting associated files when deleting a local message, but it seems that the property if a message is just plain text or is previewable is determined dynamically by parsing the content every time and not stored as a property in the DB. Is there a reason not do the latter? On the first sight that would make it easier to identify if there's a file attached to a message and to gather all media belonging to a chat w/o parsing all messages, ie implementing a media file browser.

I believe the reason is "someone ought to do it", but I may be wrong. I think a media file browser wouldn't be required if attachments were written in chat-specific directories?

nicoco: that wouldn't help to do fancy things like jump to the message for file X, in case of MUCs group files per user, ...

If that is be welcome, I'd be keen to look into that at some point in the future. ✏

If that is welcome, I'd be keen to look into that at some point in the future. ✏

Sure, a dedicated media browser could do some additional stuff that a file manager can't, but is probably 100x more work to code and maintain.

Or one uses the search feature and a plugin would do it.

> fjklp: I believed you generally send voice messages to your friends and not strangers, so the privacy concern is not really given? :) but the server also gets to know the filename ↺

> but is probably 100x more work to code and maintain. But would it really be so much code? The code for rendering previews, jumping to a message, etc. is already there. What's needed is a method to query the media files and to display them in some order in some list or tiles. Shouldn't add too much I hope

Maybe… GUI work drives me crazy usually, so I'm not the best person to talk to about this 😁️. "WHY IS THIS THING NOT ALIGNED" "WHY DOES IT JUMP WHEN I RESIZE" "WHY DOES IT SHOW A HORIZONTAL SCROLLBAR NOW" and there goes my keyboard.

Anyway, I'd surely use the media browser if there were one, but per-chat dir for attachments would already be huge IMHO

😃 Sure!

nicoco: what's left to do at your MDS MR?

maximus, cal0pteryx, there is no need for any update on a -git package in AUR as long as the build instructions don’t change, it will always pick the latest commit on whichever branch we are on.

Many users flag such packages out of date just because the description doesn’t give the latest commit, but if they build it they will have the latest one.

Link Mauve: thanks for clarifying

concerto: you need both gajim and nbxmpp to be on the latest commits

I'd like to have a light media browser as well. And yes, tracking files in a DB would be the first step

See https://dev.gajim.org/gajim/gajim/-/issues/11285

https://share.snikket.linuxkompis.se/upload/igG_-5qa3JA8cIivWMCUNhrE/gajim.png

cal0pteryx: This is far from a perfect mockup, but it shows the current spacing vs a more roomy version. :)

Ah, you mean quotes, not replies :) I don't know what it was, but there was a reason why we kept space narrow there. Maybe I remember

Sorry. I'm tired.

cal0pteryx: Let me know if you remember why you where wrong. ;)

concerto, the last debian version is gajim-nightly_20240605-1_all.deb

nbxmpp: python3-nbxmpp-nightly_20240605-1_all.deb ✏

you can check it here https://ftp.gajim.org/debian/dists/unstable/main/binary-all/

> ‎03/06/24 | 08:35:30 ‎cal0pteryx‎: bodqhrohro: legacy clients get the same experience as with xep-0393 quotes And why are there different functions (`insert_as_quote` and `quote_text`) implementing the same quoting algorithm? I see a DRY violation there.

bodqhrohro: if it's duplicated and unnecessary, MRs are welcome

🫡

It seems gajim-nightly requires python3-sqlalchemy >= 2.0.0 but I only have v1.4.50+ds1-1

concerto: yes, that's only available in debian expermental(?)

I guess I'll leave the issue be for now. Don't want to get into adding experimental, pinning, etc.

concerto: or move to Flatpak :)

Waste of space 🙁

wymm

flatpak has shared rumtimes

i'd say gajim is one of the apps worth installing as a flatpak

wurstsalat pushed 1 commits to branch gajim/master cfix: VoiceMessageRecorder: Make error strings translatable - https://dev.gajim.org/gajim/gajim/-/commit/634db69db311e32358258c34be8394fc5fd4ceca

> nicoco: what's left to do at your MDS MR? meson: I use it and it works for me with Cheogram! Not sure what’s missing. Right now it only works to sync the read state if the read message is the latest message in a chat, which is how it works with carbon displayed markers too. I suspect lovetox may want to improve that part before merging MDS

I see, thanks for the feedback!

> I'd like to have a light media browser as well. And yes, tracking files in a DB would be the first step Great, that we are on the same page :D ↺

meson: while testing the latest nightly on Windows, I see an error after recording a voice message. mic sounds are picked up fine and it shows the waveform, but when stopping the recording, I get an empty player and when trying to send, it shows nothing in the send dialog: ``` 2024-06-05T20:01:30 (D) gajim.gtk.voice_message_recorder_widget| Stopping recording 2024-06-05T20:01:30 (I) gajim.gtk.voice_message_recorder| Merging opus files started 2024-06-05T20:01:30 (D) gajim.gtk.voice_message_recorder| gerror code: 6 2024-06-05T20:01:30 (D) gajim.gtk.voice_message_recorder| gerror domain: gst-resource-error-quark 2024-06-05T20:01:30 (D) gajim.gtk.voice_message_recorder| debug: ../gstreamer-1.24.4/plugins/elements/gstfilesink.c(507): gst_file_sink_open_file (): /GstPipeline:pipeline0/GstFileSink:filesink1: system error: Permission denied 2024-06-05T20:01:30 (D) gajim.gtk.voice_message_recorder| Error when merging the recordings! 2024-06-05T20:01:30 (D) gajim.gtk.voice_message_recorder| Merging files finished 2024-06-05T20:01:30 (D) gajim.gtk.preview_audio | Could not successfully load audio. Duration is zero. ```

this is the portable version on Windows. not sure if temp file management plays into that

Good question. Can you check if it creates files in the temp folder? it should output the URIs to the files in the debug log ✏

but it sounds like it cannot write the files into the tmp folder.

meson: yes, it creates .part* files in the temp directory just fine

> gst_file_sink_open_file () but then it cannot write the final output file. Weird

I have no idea, why it works for the .part files but not for merged file, because the paths are all the same.

cal0pteryx, any anti virus interfering?

Nope

maybe run it with GST_DEBUG=4 in the hope that it gives us more hints

Permissions seem to be fine

meson, you shouldn’t need to have a file in the end, you could stream it to the HTTP server or Jingle recipient as is from Gstreamer.

Sure, a file is needed to fit the use case and client ecosystem

Link Mauve, but I feed the file first to the audio player and the recording can be continued afterwards before sending ✏

Ok.

Link Mauve, but if you have any idea, this is the part where it stumbles: https://dev.gajim.org/gajim/gajim/-/blob/master/gajim/gtk/voice_message_recorder.py?ref_type=heads#L468

My knowledge of Windows is pretty much non-existent.

When running from source, it works fine. Though the temp dir is different, it's C:\msys64\tmp\gajim-***

I still find this whole opus enc/dec/enc dance to be quite terrible.

You should always keep the data in highest quality until the final encode.

:D but I didn't want to write the uncompressed raw stream to disk, it could occupy quiet amount of space

You could use actual ramdisk locations if you want to avoid using the disk, on Linux the best location would be XDG_RUNTIME_DIR I think.

> highest quality until the final encode. but does it make any difference for simple mic voice recording?

I would expect so.

I will keep it in mind for follow-up improvements :)

Link Mauve, we are a cross platform client, the benefit from implementing platform specific behavior needs to be major ✏

is there a way to change the xml console colors? my eyes are itching

lovetox, there are ramdisks on all OSes I would expect.

And I also would expect GLib or Gio to abstract that somewhere.

cal0pteryx, does the issue happen on the first stop of the recording or on the 2nd onwards?

meson, i would tip that this is a garbage collection problem

maybe the tempdir is cleaned up for some reason while merging

The .part files persist when pausing the recording

meson: does not matter

First stop, following stops are alike. Files get created just fine. ✏

meson i see in the code VoiceMessageRecorderButton and you connect a destroy event and cleanup the voicerecorder afterwards

but under which conditions is VoiceMessageRecorderButton destroyed?

on shutdown of Gajim?

> on shutdown of Gajim? I think so ↺

otherwise, it persists afaik

Direct recording yields the same issue

it's instance is created with the message action box

nope, it does not get called on quit

i think this is dead code

which means also temp files get never deleted

That's bad. I will be back in hour.

so you potentially run into problems when you access already existing temp files

Looks like it, quitting gajim does not clear the temp dir

lovetox: a random folder name is chosen on start

gajim-random/voice-message-timestemp.ext.part*

how does gajim determine whether a client on a server has the right to upload? what does the xml query look like?

Nepptün: https://xmpp.org/extensions/xep-0363.html#disco

> nope, it does not get called on quit What's odd, because it looked like folders got cleaned up on my Tumbleweed system. cu later.

Seems I was wrong: the temp folder gets deleted when quitting gajim

(maybe some tempfile package auto cleanup)

yes > when the object is garbage-collected or during interpreter shutdown

ok, interpreter shutdown it is. those part files persist until I Ctrl+C gajim

cal0pteryx, you say it works from source on windows?

Yes

But it uses a different temp folder if run from source

> maximus, cal0pteryx, there is no need for any update on a -git package in AUR as long as the build instructions don’t change, it will always pick the latest commit on whichever branch we are on. I see but, I thought it was `gajim-git` was basically equivalent to `gajim-nightly` on flatpak, so basically gajim on branch master, but I don't get any updates even though commits are being pushed everyday ↺

thats weird, where is the tmp folder when used from the portable version?

maximus: you need to explicity reinstall I think

lovetox: C:\Users\X\AppData\Local\Temp

(where X is my user)

When running from source, it's C:\msys64\tmp\gajim-***

and it only happens on sending, it creates the parts normally?

It creates parts normally, in the respective folder, yes. Once I stop the recording, it tries to merge the parts. That's where it fails

cal0pteryx, ohh thank you very much, it did the trick!

I thought they automatically gets updated on every push

cal0pteryx, on the very first start/stop after start of the client?

meaning it never works under any condition?

cal0pteryx can you try it with a simpler pipeline? ✏

lovetox: yes, it never works when using portable, regardless of how many times I stop the recording

meson: sure

Reproducing is hard though, since I can't edit the portable install, and it works fine when running from source

Replace it with `audiotestsrc num-buffers=100 ! audioconvert ! audioresample ! filesink location=self._filepath` or something similar

bodqhrohro: > 🫡 oink ↺

cal0pteryx: > bodqhrohro: if it's duplicated and unnecessary, MRs are welcome oink oink ↺

cal0pteryx, you can move the source file for the recorder into the folder, then simply edit it ✏

Oh, neat!

cal0pteryx: to build gajim portable, executing win/dev_env.sh is enough or is more needed? There's now a gajim.exe in _build_root/ucrt64/bin/ but I get an error when I try to execute it from the Windows explorer

meson: build.sh, and rename gajim.exe to something different

Build failed, it couldn't find Python.h

Trying it again after pacman -S python-devel

``` https://conference.gajim.org:5281/pastebin/ec10f502-3e4f-4714-a859-a4ae0fc60458

No idea why I prefer developing under Linux :)

meson: not sure where that comes from. I'm building the installer using mingw64

gajim.exe (the setup) is created in win/_build_root/gajim.exe

running win/build.sh from a mingw64 terminal

Not sure where I would put that simpler pipeline you mentioned, meson