🖥-vesktop-development

[floral ledge]

I ALREADY see another issue

They dont follow the spec again

[humble mortar]

what is it

[floral ledge]

Ok so the spec is like

[humble mortar]

kde or hyprland?

[floral ledge]

KDE

You can only register shortcuts once in a session

According to the spec

In KDE you can just do it however often you want

[humble mortar]

uhh no?

[floral ledge]

I do not see the prevention in here nor in setactions

I also found your duplicate issue

setactions has no safeguard against duplicate either

[humble mortar]

that was an intentional choice theres a comment in the impl pr

where in the spec does it say once a session

[floral ledge]

Okay well its a bad choice because hyprland only allows once per session

This isnt a spec if no one follows it

Or only some people follow it

[humble mortar]

that'd be a hyprland issue I think

no where does it say once

[floral ledge]

yes it does

I vividly remember this from the spec

@humble mortar

[floral ledge]

I feel like relying on applications to check the DE/WM just to correctly implement shit is a terrible idea

[humble mortar]

how have I never seen that 😭😭

[floral ledge]

This is a case where an application would or it would have to take the more strict behavior of destroying the session

to reregister

on every de/wm

[floral ledge]

Oh yeah it gets worse

https://invent.kde.org/plasma/xdg-desktop-portal-kde/-/commit/e19e17a7892d76b9afa0a6ec50f3f02dddac50bb

They actually removed the shortcut clearing

[humble mortar]

ok so why did the kde implementor do that

[floral ledge]

the reason its there is the older version of the spec

accepted the shortcuts in CreateSession

which obviously.. is once per session

https://invent.kde.org/plasma/xdg-desktop-portal-kde/-/commit/a748343ee8a8cdfaeaa4a40544ce41a5e3b56e97

Essentially the implementor just Moved It and didnt read the spec

Saw changed to listshortcuts and just moved it in 5 seconds

[floral ledge]

Entirely unmentioned in the merge request description or reviews

[humble mortar]

this is so cooked

so hyprland correctly does it yes?

[floral ledge]

yes

lmfao

ohh my god

this pisses me off even more

[floral ledge]

this "only once" line was added in a pr with the message

global-shortcuts: Improve documentation
Various small cleanups in wording, formatting, and whatnot.

that is NOT a small cleanup and improvement that is a SPEC CHANGE

https://github.com/flatpak/xdg-desktop-portal/commit/737c493dc8dbbbb0d2cdab0bd7e7f5ab319e5e8a

[humble mortar]

wait so is the spec wrong

[floral ledge]

no

the spec is correct

its just a stupid commit msg

[humble mortar]

I guess

hm

guhhh

[floral ledge]

What happens if you pass a shortcut with some stupid fucking LOOONG description

Throw like 2kb of @'s in a shortcut description please i wanna see that

[humble mortar]

I'd assume itd just error out at the dbus level

[floral ledge]

and i dont have a plasma rig to test

nuh uh

max dbus message length is 128mb

[humble mortar]

my laptop is off I was about to go to bed 😭😭😭

[floral ledge]

aww

Input validation is not in the global shortcuts spec yet and no one validates anything

What could possibly go wrong

[humble mortar]

well atp the app would be intentionally breaking things and that'd count as malware

so like who caress

[floral ledge]

No but heres the idea

It saves in system settings

Can you just crash loop kde lol

Super annoying to fix!

[humble mortar]

last time an entry got corrupt (the no Id thing) restarting kde cleared it

[floral ledge]

real

still no input validation is bad too

[humble mortar]

we will see

!remindme 11h attempt to cook kde

[small lion]

Alright @humble mortar, in 11 hours: attempt to cook kde

[humble mortar]

I'll make an issue for the bind return thing then too

maybe you should make an issue for the non spec conformity

https://invent.kde.org/libraries/xdg-portal-test-kde/-/blob/master/src/xdgportaltest.cpp#L236 ha

[sour mural]

xdgportaltest.cpp: Line 236

{ QLatin1String("session_handle_token"), "XdpPortalTest" },

[humble mortar]

their test impl uses a static session handle token

[floral ledge]

oooh my god

Like

The spec isnt great

But they could at least try to read it??

The spec has been reasonable for literally a year

thats when all the fixes were made to it

[humble mortar]

this is their test impl for the entire portal

are WE interpreting it wrong

[floral ledge]

no

Every single client library uses it correctly

LOL

[humble mortar]

I'm beginning to go insane

[floral ledge]

ashpd libportal both do it correctly

[humble mortar]

if libportal does it with a random then that has to be right

so kde is insane

[floral ledge]

Unless the spec is saying something in clear english where it does not in fact mean the clear english of "use a library prefix combined with a random number"

[humble mortar]

😭

[floral ledge]

Then it is definitely kde's fault

[humble mortar]

where's the spec on how it grabs the appid

if their test app is made to conform it'll be the random token issue all over again because it doesnt have a desktop entry & its not launched from it

why are app icons and stuff implemented different from the session appid grabbing

this is all insane

[floral ledge]

You dont need one

Implementation detail

All you're guaranteed is the app id is:

• most likely in a format defined by https://systemd.io/DESKTOP_ENVIRONMENTS/, and
• that retrieved app ID as defined by the systemd spec most likely/does conform to the spec here https://developer.gnome.org/documentation/tutorials/application-id.html except for the 'two or more elements' bit

It grabs the appid just based on the systemd desktop environments spec for a systemd unit

For flatpak and snap its some insane garbage code I haven't read

As an aside the gnome 48 feature freeze is on February 1st

They have 21 days to get the global shortcuts pr merged into xdp-gnome

for support on day 1 of gnome 48

[simple grove]

litemotiv

[Vencord/Vesktop] New comment on issue #1045: [Feature Request] Overlay scrollbars

maybe use flatpak run and edit the .desktop to use that?

I'm just not sure if the parameter is processed by Vesktop, i've used the approach suggested by @Vendicated and the runtime shows that's invoked with the parameter:

This does however not affect the scrollbars inside Vesktop, so it appears that the parameter doesn't propagate as expected.

I'm not sure how else i could approach this.

[humble mortar]

If no application ID is available, the launcher should generate a reasonable name when possible (e.g. using basename(argv[0])). This name must not contain a - character.

Should kde/hyprland not do this instead of falling back to the token or nothing

[simple grove]

Covkie

[Vencord/Vesktop] New comment on issue #1045: [Feature Request] Overlay scrollbars

does the flag work on chromium and discord.com/app

[simple grove]

litemotiv

[Vencord/Vesktop] New comment on issue #1045: [Feature Request] Overlay scrollbars

does the flag work on chromium and discord.com/app

Hmm good point, it appears the scrollbar always stays visible in the main panel there as well (non-overlay).

In Firefox the scrollbar in the main panel is overlay, so perhaps there is custom scrollbar code specifically in Chromium/webkit CSS that is forcing the scrollbar to stay visible.

[small lion]

@humble mortar, <t:1736578156:R>: attempt to cook kde

[simple grove]

MochiVN634

[Vencord/Vesktop] New comment on issue #1032: [Feature Request] Persistent stream settings

Yes, please, this is absolutely necessary, I often stream to my friends in 1080P 60FPS and every single time I have to switch it from 720P 30FPS.

[brazen kite]

now who wants to do the honors of testing the PR? https://gitlab.gnome.org/GNOME/xdg-desktop-portal-gnome/-/merge_requests/166

[simple grove]

littlecube23

[Vencord/venmic] New comment on issue #34: No sound gets shared in screenshare

I also have this issue on Zorin OS.
Settings selected when screensharing

How audio routes as shown by Helvum

What I have to do to get audio to work

OS: Zorin OS 17.2 x86_64
Kernel: 6.8.0-51-generic
stable 35801...

[floral ledge]

yeah ive seen the pr

its just not done yet\

[brazen kite]

I doubt it'll be merged by then

there needs to be at least 3 years of bikeshedding for a big feature in gnome to be merged

[floral ledge]

And it depends on https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3343 and https://gitlab.gnome.org/GNOME/gnome-control-center/-/merge_requests/2485

the gnome-shell one is literally a edited js file

and a new xml file

and it has still not been merged after 7 months

[floral ledge]

this isnt even a big feature either

[grizzled hemlock]

@humble mortar sorry for the ping but is vesktop screenshare being very choppy on windows a known issue

quite a few people have mentioned it in support and i can reproduce it on my laptop

[humble mortar]

is it choppy on the streamers client or just the viewer

[grizzled hemlock]

just the viewer

[humble mortar]

disable hardware accel

ive noticed and reproduced hardware encoded streams are COOKED on the viewers end

except on web clients

and mobile

something is going wrong between webrtc and whatever custom solution discord desktop uses

[grizzled hemlock]

@humble mortar just did that, quit vesktop from the system tray and its still really fucking choppy

[humble mortar]

idk then

stream from chromium to a web client to confirm if its just vesktop

ive seen people say it happens on other custom web based apps so I dont think it is just us

also try to view vesktop from chromium

[grizzled hemlock]

I'm already watching from a web client, so I'll try streaming from one as well

[humble mortar]

its choppy when viewing from web on windows?

[grizzled hemlock]

it was choppy when streaming from a vesktop client on windows and watching from a web client on linux

i will now try streaming from web on windows to and watching from web on linux

@humble mortar its choppy when streaming from web

on chrome

[humble mortar]

is the stream being hardware encoded

[grizzled hemlock]

how do i find that out

[humble mortar]

idk I dont use windows 😭

either way its something wrong with chromium

maybe try to stream from firefox

[grizzled hemlock]

alr

[humble mortar]

if its laggy viewing that too discord cooked their native module

[grizzled hemlock]

huh wdym

isnt there no native module on web

[heady merlin]

yeah there isn't

and that why it sucks ass

[humble mortar]

it could also be a bandwidth issue

discords native module uses quite a lot less

[grizzled hemlock]

im on the same wifi as my desktop

i should be fine

[humble mortar]

¯_(ツ)_/¯

[grizzled hemlock]

@humble mortar its perfectly smooth on firefox

chromium cooked

[humble mortar]

try older versions

go find the regression 🔫

[grizzled hemlock]

too lazy

[humble mortar]

booo

actually pause

was Firefox hardware encoding

[grizzled hemlock]

idfk

[humble mortar]

😭😭😭😭😭😭😭

[grizzled hemlock]

i dont stream on windows so i cant be bothered to find out

if you know how to check ill do it

[humble mortar]

idk check process viewer gpu usage?m

task manager

also this is def a larger issue because vesktop/chromium hardware accelerated stream -> macos desktop client = didnt load but -> macos web client (chromium based) it loaded and was fine

does chromium -> firefox work fine

[grizzled hemlock]

huh

chrome was choppy

firefox wasnt

[humble mortar]

stream from chrome to firefox

maybe the upstream libwebrtc made a change to their impl and discord hasn't pulled them in to their fork

[grizzled hemlock]

huh why would that matter

wouldnt there be no native module in browser -> browser

[humble mortar]

discords native module is patches ontop webrtc

[humble mortar]

if streaming from chrome to Firefox works fine this is almost likely the issue

[brazen kite]

it's a big feature for gnome 😱

[simple grove]

padoruuuu

[Vencord/Vesktop] New comment on issue #1014: [Bug] Significant performance drops while streaming

i have this issue on sway

[simple grove]

Covkie

[Vencord/Vesktop] Issue closed: #1014 [Bug] Significant performance drops while streaming

Covkie

[Vencord/Vesktop] New comment on issue #1014: [Bug] Significant performance drops while streaming

please use online resources such as https://wiki.archlinux.org/title/Chromium#Hardware_video_acceleration for how to get hardware acceleration (encode and decode) working on Linux.

[humble mortar]

@floral ledge do you think I should make a single issue on kdes bug tracker for these or one for each

[simple grove]

ArcPlayz

[Vencord/venmic] New star added

[simple grove]

Curve

[Vencord/venmic] New comment on issue #34: No sound gets shared in screenshare

I also have this issue on Zorin OS.
Settings selected when screensharing

How audio routes as shown by Helvum

What I have to do to get audio to work

OS: Zorin OS 17.2 x86_64
Kernel: 6.8.0-51-generic
...

[floral ledge]

I cant figure whether the developers will be more pissed at separate issues than the one

I think you make the one to save yourself 15 minutes

[floral ledge]

the side note is less of an issue

the spec does not specify how you should save shortcuts

[simple grove]

ibanks42

[Vencord/Vesktop] New comment on issue #887: [Feature Request] mute/unmute with a command

Here is a script that doesn't require setting PULSE_PROP="discord.id=vesktop" vesktop. It is also modified to not mute screenshare, and allows adding different Vencord clients like Equibop.

#!/bin/bash

DISCORD_ID="io.github.equicord.equibop"
WANT_MUTE=$1
DISCORD_IDS=("dev.vencord.Vesktop" "io.github.equicord.equibop")

APP_ID_SELECTORS="("
for i in "${!DISCORD_IDS[@]}"; do
  APP_ID_SELECTORS+=".info.props.\"pipewire.access.portal.app_id\" == \"${DISCORD_IDS[$i]}\""
  i...

[humble mortar]

but it does define how to get the appid no?

[floral ledge]

No

LOL

The spec doesnt actually define anything about the app id

https://github.com/flatpak/xdg-desktop-portal/blob/main/src/xdp-utils.c#L291

[sour mural]

xdp-utils.c: Line 291

xdp_is_valid_app_id (const char *string)

[floral ledge]

The portal source code defines it

[floral ledge]

https://github.com/flatpak/xdg-desktop-portal/pull/1521

THANK FUCK

they need to do this

[simple grove]

littlecube23

[Vencord/venmic] New comment on issue #34: No sound gets shared in screenshare

The settings before I hit stream

Stream Settings

What shows in QjackCTL

Friends cannot hear audio.

Also some more info just in case?
Server Name: PulseAudio (on PipeWire 1.0.7)
Server Version: 15.0.0

I wo...

[brazen kite]

I think that's a better approach

especially if it isn't sandboxed

[humble mortar]

kde instantly +1'd holy moly

no movement past that tho 😭

btw https://bugs.kde.org/show_bug.cgi?id=498575

[floral ledge]

also

https://github.com/flatpak/xdg-desktop-portal/discussions/1564#discussioncomment-11814091

I made this discussion

maintainer agreed it would be good to point out in the spec

[simple grove]

Covkie

[Vencord/venmic] New comment on issue #34: No sound gets shared in screenshare

Disable microphone workaround?

[floral ledge]

https://github.com/flatpak/xdg-desktop-portal/discussions/1566

New discussion

[humble mortar]

conflicting things are up to the implementation

kde has unset shortcuts

hyprland wouldn't since its all config driven

trigger_description is however the implementation conveys how to trigger a shortcut

so it obviously won't have unset info

thats for the settings panel

shortcuts are already 'saved' for relaunches based on the implementation

[humble mortar]

this makes sense to have in the spec though

kde does appid then prefixed session token

hyprland does appid then nothing

soo

actually I bet theres a spec for saving persistent session stuff

which covers everything

[simple grove]

littlecube23

[Vencord/venmic] New comment on issue #34: No sound gets shared in screenshare

Proof With Microphone Workaround Turned off

Before I hit stream

What QjackCTL shows

Friends can still not hear. Ive verified I am not muted on their end and their stream volume is still up

[humble mortar]

wait no its literally defined to be app_id

[floral ledge]

no

kde just registers the shortcuts

[floral ledge]

where

[floral ledge]

and then it breaks due to a bug

[humble mortar]

it doesnt break it just misreports

its registered and unbound

but still says ctrl+1 in the trigger_description which is wrong

[humble mortar]

on any impl spec for CreateSession iit has the app_id field

[floral ledge]

yes

it doesnt specify how shortcuts should be saved

remember kde was doing with appid + token

[humble mortar]

that was a mistake i believe

[floral ledge]

yeah but still spec doesnt specify

[humble mortar]

its implied like that other thing

[floral ledge]

this is notr implied

this will break compatibility depending on program behavior

between backends

[humble mortar]

trigger_description is just a string

the app should treat it as such

and display it to the user

its not something to be parsed or anything

[floral ledge]

okayu

that doesnt answer the question

what do you do when theres a conflicting keybind

do you not register it?

mind you theres multiple possible answers

[humble mortar]

its up to the impl

its preferred_trigger for a reason

[floral ledge]

thats not an answer

one app expects it to do one thing

other app expects another

[humble mortar]

you dont rely on what you send

[floral ledge]

does the app just not know what the trigger key is

[humble mortar]

you rely on what it returns

[floral ledge]

you cant figure out what the trigger key is

[humble mortar]

trigger_description is what you rely on

and you display it to the user

[floral ledge]

gg

[humble mortar]

i dont get why thats an issue for you

[floral ledge]

so you register and return "unset" from the implementation

or some arbitrary value

implementation defined

[humble mortar]

you ask to register to ctrl+1

the impl returns what to show the user in trigger_description

simple

[floral ledge]

that does not answer my question

you keep saying that but that doesnt answer my question

what should an implementation do with conflicting keybinds

you cant just say "up to the implementation" because im asking what the implementation should make up

[humble mortar]

disregard or store it like kde

hyprland doesnt care about preferred_trigger at all

kde stores it as 'default shortcut' but unchecked if its conflicting

[floral ledge]

I think the reason the functionality was hard to think about for me is because I forgot we should just implement it completely differently in vesktop for wayland/xdg compared to any other platform

we should just be registering all the possible discord global shortcuts with no preferred trigger

on vesktop start

[humble mortar]

yeppp

[floral ledge]

and then the user can configure it

[humble mortar]

its a different paradigm

[floral ledge]

Yeha

[humble mortar]

the keybind ui will simply be a display

[floral ledge]

I forgot that I should have had that paradigm in mind

The keybind UI shouldnt even show shit

Just turn it off on wayland except for the like help thing that shows default ones

[humble mortar]

well it should still be there to display the trigger_description

https://github.com/tuxinal/venbind/pull/1#issuecomment-2585930319 i laid it out here

[floral ledge]

Yeah this is true

Also that PR is ready

So is the windows one

Basically they're both like. Not actual full implementations

I am going to have to clean up both afterwards

theres literally no keyup support in either

I need to clean up the key trigger logic in both

Including the registration logic

the tomorrow in question

[humble mortar]

lol

i'd help if i was confident in my rust abilities lol

[floral ledge]

I would help but I'm waiting for stuff to get reviewed and merged

before I start overhauling things

[humble mortar]

just overhaul and tell tuxinal to cope tro

[floral ledge]

I need the windows pr and xdg pr in to start overhauling

[humble mortar]

merge locally

[floral ledge]

I could do that in my own branch but thats really annoying

[humble mortar]

the git rebase in question

[floral ledge]

if the xdg pr changes at all before being merged

the rebase will be a disaster

[humble mortar]

his last comment on it implies he was gonna merge as is

sooo

also remember an internal debounce thing

i dont think any keybinds discord has are press and hold incremental ones

[floral ledge]

yes the keydown spam wont work for any discord keybinds

will be fixed also

[humble mortar]

yep yep

[brazen kite]

🙏

you're doing god's work

[humble mortar]

gods work is going insane with freedesktop standards

[scenic hollow]

i was waiting to see if the other contributor responds and undrafts their pr

i guess i could just undraft it myself

oh i just saw your responses

[floral ledge]

@scenic hollow conflicts resolved

i regenerated the lockfile

i plan on porting to napi-rs so CI will be fixed soon with new pr

also i think refrain from making a new version until this is somewhat usable software

[scenic hollow]

uh also we do kinda rely on the fact that you can define multiple keybinds per session
idk how else we would do it? how would the user set up a new keybind in one session?

[floral ledge]

okay other thing i was curious about

is there something that stops us from using wayland seats

instead of xdg

per-compositor security?

[scenic hollow]

does that not require focus on the window?

[floral ledge]

I mean not according to the Wayland protocol

I think that'd be implementation specific

"A seat typically has a pointer and maintains a keyboard focus and a pointer focus."

hmm

[scenic hollow]

Using the wl_seat.get_pointer request, clients may obtain a wl_pointer object. The server will send events to it whenever the user moves their pointer, presses mouse buttons, uses the scroll wheel, etc — whenever the pointer is over one of your surfaces

[floral ledge]

where is this

[scenic hollow]

next page

[floral ledge]

its in the protocol appendix?

isee

[scenic hollow]

https://wayland-book.com/seat/pointer.html

[floral ledge]

oh

thats the book

im talking about the spec mainlty

anyways yeah, that's true

I can see in hyprland they check focus before sending any events

total security model victory

But it technically is up to the compositor

You could make a shitty compositor with no seat security

LMAO

[scenic hollow]

yeah and ask every vesktop user to use it if they want shortcuts and then support nothing else

[floral ledge]

I lied

defined for wl_pointer in the spec

"The wl_pointer interface generates motion, enter and leave events for the surfaces that the pointer is located over"

nevermind i didnt lie

god this is confusing

I cannot tell where the security is actually defined lmfao

[scenic hollow]

either ways i'd assume it's not intended to be used as a way to define shortcuts

[floral ledge]

Found the security in hyprland

they have grabs

This is the actual security

1 active seat grab at a time

[floral ledge]

and it only sends to seats made by the surface client

Well the security model is much better than windows

good luck secretly keylogging ig

[floral ledge]

read evdev and ask every vesktop user to run vesktop as root

gg

[floral ledge]

what do you mean

does discord support one keybind for multiple keys

like one action

ok yeah it does

You don't

You would have to end and recreate the session really

the idea is this supports 1 shortcut -> 1 keybind trigger

and you remove the ability to change them from the settings panel. you just display the trigger_description and you change them from the compositor's method of changing them

on kde being system settings, hyprland being the configuration files

[scenic hollow]

i meant we currently call BindShortcuts once for each shortcut defined. according to the spec we can't do that

An application can only attempt bind shortcuts of a session once.
so we'd have to first gather all our shortcuts and then add them in one BindShortcut call. idk how we'd do that

[floral ledge]

keyword

"of a session"

free the session

make a new session

[scenic hollow]

ok but that's kinda crappy in it's own way

[floral ledge]

That is ONLY if we want to support the user adding more keybinds though

[floral ledge]

The correct way is whenever Vesktop starts we grab all the possible actions that can be bound

since there's only like 15 of them

that discord allows you to bind

And we bind them

No preferred triggers

Edit the keybinds through your compositor

Basically the paradigm is going to be completely different to any other platform for how it will work

[floral ledge]

I don't know if anyone actually seriously binds an action to more than one key very often at all

Seems very niche to go through the effort of session reregistration to support unfortunately

And if they really want that it should be a compositor supported thing

AKA Hyprland supports that

Just add another config entry

With the same shortcut

Boom the hyprland event dispatcher will handle it for you

idk about kde

[scenic hollow]

kde disables shortcuts on conflict

[floral ledge]

That's not what I'm saying

I'm talking about binding several keys to one action

I imagine KDE just supports 1

[scenic hollow]

OH yeah i read that backwards

[floral ledge]

yeah

[scenic hollow]

no i think kde supports multiple actually???

[floral ledge]

how

can u like add more in the settings panel

lol

[floral ledge]

Binding one key to multiple actions is flat out never happening

Dead on arrival because no compositor is gonna handle that properly

Except hyprland lawl

[scenic hollow]

yeah you can just checked

[floral ledge]

KDE just breaks whenever theres a conflict its stupid

[floral ledge]

If it does happen then its the portal backend's issue not ours at least

dont have to worry about it

[floral ledge]

seems like hyprland's config file is the most solution free issue

and they dont have to spend time making 4 prs to implement it in UI

they're on to something here!!

Effortless multiple keys to one action and multiple actions to one key !

[scenic hollow]

honestly in this scenario i'd rather ask the xdp people to consider changing the spec cause that's a weird requirement

[floral ledge]

It's not really a weird requirement

Because you're seeing this from a Vesktop perspective

Where all of this requires some insane JS modding and app modification

When you're like making a normal desktop app and you want shortcuts you have a list of actions you want to be able to be bound to a key

So its actually really easy for that use case

All i do is i pass the portal my actions and boom

The user can configure shortcuts in the compositors' method for that whether that be a UI or a file and I dont have to do any input logic

I just get my activated and deactivated events

And the compositor can support multiple keys to one action and multiple actions for one key on its own in UI/configuration

It's pretty easy/elegant for the vast majority of an application's purposes

[scenic hollow]

guhh yeah i guess

[floral ledge]

It's a different paradigm to discord and like conventional applications though where they have hteir inbuilt settings UI and you can go and do whatever keybinds you want in there

But that's part of the security model

All of that is handed off to the compositor

So you really don't actually want an application to be abusing preferred_trigger and freeing/creating sessions

The correct ergonomic way is a session at app start and sending over all the possible shortcut actions

And the compositor handles the rest

Frankly it's easier than windows input where you have to go use some library that makes a hook or some insane garbage and receives every input event on the entire system and has to filter it

If not for d-bus requiring a wohle lotta boilerplate to interface with at least

but its just IPC instead of insane global input hooking

[humble mortar]

tbh just steal/remake the the shortcut ui

patching it for our uses will be insane

[floral ledge]

Yeah the idea I think is literally rip out like the entire shortcut ui

[humble mortar]

literally yoinking it would be liscence violation I think

[floral ledge]

Like just rip out the top part not the keybind help at the bottom for the default keybinds

yeah

don't yoink it just remake it

[humble mortar]

yop

[floral ledge]

It would not look good anyway with the changes

You'd have to remove the record button

Then display the trigger description (which on hyprland is literally always an empty string mind you)

[humble mortar]

2 billion patches vs writing a bit of ui

[floral ledge]

yes

I think you can just remove the react element at the top and then replace it

tbh

[humble mortar]

it doesnt return anything even if you set a bind in the config??

[floral ledge]

It does heres the trigger_description code

combines all the trigger names

[floral ledge]

Nope lol

Too lazy ig

[humble mortar]

open bug report

[floral ledge]

Not a bug implementation decision

[humble mortar]

thats not spec conformant

[floral ledge]

That'd be a feature request

[humble mortar]

the spec says it should return a string on how to trigger

[floral ledge]

"User-readable text describing how to trigger the shortcut for the client to render."

[humble mortar]

exactly

[humble mortar]

kde is 1 bind per shortcut

[floral ledge]

I don't think it's super duper easy for hyprland to pull this off

[scenic hollow]

you can add more in the settings page

[floral ledge]

Their wayland protocol doesnt support that

[humble mortar]

query whatever config constructor thing they have for the appid+shortcut id

[floral ledge]

Yeah thats not simple

[humble mortar]

yes but multiple shortcuts can't have the same bind active

[floral ledge]

The portal interfaces only through their wayland extension

https://wayland.app/protocols/hyprland-global-shortcuts-v1

[humble mortar]

shortcuts can have multiple binds

[floral ledge]

They would have to modify their protcol to support returning the triggers

And signals for when those change

It would be such a mess

[humble mortar]

they implemented GlobalShortcuts ontop of their own portal?

insane

[scenic hollow]

I READ IT BACKWARDS AGAIN sorry

[floral ledge]

how else should they do it

They made a wayland protocol extension for global shortcuts specific to hyprland

the portal talks to the compositor through their extension

everyone is happy

[humble mortar]

insane layers

[floral ledge]

No other way to do it unfortunately

[humble mortar]

😪

[floral ledge]

KDE does it through QActions or some stupid shit

I don't know how kde works lol

I dont know why it uses qt

Nor how the compositor registers them

[humble mortar]

kde IS heavily reliant on qt things

very INSANE

[floral ledge]

I can tell

How the hell does that even work

Do they have a custom fucking layer on top of qt that makes shit actually work

like

[humble mortar]

qt isnt solely ui

[floral ledge]

How do those qactions work in the compositor lmfao

I know it's not solely ui

Qt has an insane amount of stuff in it

Qt sucks every app that uses it takes 200mb of memory now I might as well be using an electron app

[humble mortar]

anyway yeah open a report on hyprland

let them figure out how to mangle their shit

[humble mortar]

I think thats just the specific app using an insane amount of widgets

[humble mortar]

the spec defines trigger_description as something you can rely on

so it shouldnt always be empty

[floral ledge]

https://github.com/Nheko-Reborn/nheko

chat does it suck

...

[humble mortar]

chat clients are always gonna be 'heavy'

so many things if you want it to be stylized and not simple text boxes

[floral ledge]

This is true considering halloy uses 220mb of memory on start and that's made with iced

[humble mortar]

also exe so its all bundled and can't share the libs

[floral ledge]

so fair point

[humble mortar]

also why doesnt the spec have any method to trigger opening the shortcuts settings beyond BindShortcuts

seems like an oversight

[floral ledge]

node_bindgen was the single biggest architectural mistake venbind made i gotta fix this

holyy

i have nodejs in path stop downloading it my guy

[humble mortar]

[floral ledge]

Also this is fair enough

https://github.com/hyprwm/xdg-desktop-portal-hyprland/issues/312

what is even happening here

is this lifetimes coming to haunt me

[scenic hollow]

huh no i remember having a lot of issues with u64 in general

i am removing the display_id and window_id btw for now disable the node feature

or

wait for me to commit my stuff

[floral ledge]

okay I will just revert my conversion to a move closure

and push the other windows compilation fix

[floral ledge]

Like entirely?

I will have to remove them from windows start_keybinds_internal

that or I wait for you to commit and rebase

(again)

if its entirely I can just remove it really easiyl from the windows impl

[scenic hollow]

yeah entirely i thought it's required for wayland but apparently not

[floral ledge]

alr

windows pr is good to merge then

[humble mortar]

wait does hyprctl globalshortcuts not even return what theyre bound as ?

[floral ledge]

No lol

Fairly sure not at least

[humble mortar]

thats so silly

[floral ledge]

It might

[humble mortar]

🧑‍🦯

blind leading the blind

[floral ledge]

It intentionally is that way for hyprctl globalshortcuts

Hyprctl globalshortcuts basically just gives you the info you need to add that to the config file

If you use hyprctl binds or whatever it will show up

just read the source

[humble mortar]

in not reading source on my phone 😭

[floral ledge]

I meant I just read it

sorry

[humble mortar]

I LOVE ENGLISH

READ AND READ

so good

[floral ledge]

hyprctl binds - lists active keybinds
hyprctl globalshortcuts - lists active global shortcuts solely for the purposes of easy reference/addition to the config file

it just lists the appid/shortcut description practically

hyprctl binds will list global shortcuts you actually bound

dont ask me why

[humble mortar]

so slap whatever binds is doing into the portal

unless its parsing the config

[floral ledge]

you cant

binds is reading from the internal keybind manager m_vKeybinds array

hyprctl is built into hyprland

the portal is a separate source tree that only communicates through the wayland protocol extensions

[humble mortar]

ughh

[floral ledge]

:3

I love... MODULARITY!!!

Ironically hyprland is the most reasonable and easy to interpret implementation which is hilarious

Instead of insane Qt bullshit its just a wayland protocol extension

[humble mortar]

are kde and hyprland really the only 'full' implementations

[floral ledge]

Ohh

I remember how kde's one works

Oh my god lmfao

They just create a qaction and then stick it into KGlobalAccel

Great work guys

[humble mortar]

not really that much different than hyprland tbh

tacking it onto the existing shortcuts impl

archwiki has dde as an implementation of GlobalShortcuts

[floral ledge]

no

hyprland's wayland protocol extension thing is great

super easy to find the documentation and its all specified

want to use it? u can use that protocol in any app without dbus

[humble mortar]

https://github.com/linuxdeepin/xdg-desktop-portal-dde/blob/master/src/globalshortcut.cpp

its a stub

[floral ledge]

just using the protcool with wayland-scanner or hyprland-wayland-scanner

[humble mortar]

good job archwiki guys

[floral ledge]

kde's implementation?

yeah. we use kglobalaccel

which is like ok whatever

works ig

except theres even more layers

it goes portal -> kglobalaccel -> kde daemon

so kglobalaccel is just a middleman

oh my god

wait its even worse

ohh nooooo

@humble mortar
Client dbus -> xdg-desktop-portal -> impl dbus to backend -> KGlobalAccel -> kde daemon THROUGH ANOTHER DBUS

hyprland uses a wayland protocol. they use another fucking dbus

:-(

https://github.com/KDE/kglobalaccel/blob/master/src/org.kde.KGlobalAccel.xml

[humble mortar]

youre hurting yourself

implementation on implementation

[floral ledge]

You thought hyprland's wayland protcol was bad

Yet thats a super simple portal backend -> compositor direct protocol that I can find just by going on the wayland protocol browser

on the pretty docs page

KDE? Extra middleman and an extra dbus protocol

enjoy

Its not even to the compositor

[humble mortar]

counterpoint: plasma has wobbly windows

[floral ledge]

its to kde daemon

What the hell does kde daemon do

Do they just have a service that runs and grabs input directly

do they even bother with wayland input

KGlobalAccel is so ancient I would imagine not

Oh god

The kglobalaccel source says the relevant handler is in kde daemon

(The relevant handler is not in kde daemon)

https://tenor.com/view/g-man-half-life-gif-26268160

Where is it going

No way to tell since its a random dbus request with no docs so it doesnt tell you

Where is John KDE to consult on this matter

You cant search an org for code on gitlab

[humble mortar]

clone everything locally

only solution

wait pause

[floral ledge]

I found it

They moved it from kde daemon to plasma

[humble mortar]

github mirrors exist

[floral ledge]

yeah

i found it on the github mirror

LOL

by org searching

[humble mortar]

love

I'm gonna go to bed now

you have fun..

[floral ledge]

https://www.youtube.com/watch?v=ciL7OKWSu10

my reaction

KDE devs are allergic to spacing

[simple grove]

Psychloor

[Vencord/Vesktop] Issue opened: #1047 [Feature Request] Remember last position on startup (Windows)

Discord Account

psychloor

Motivation

Whenever Vesktop starts up, it doesn't remember last position it was on my desktop unlike discord.
have to drag it to the secondary monitor every time

Solution

remember last position on drag and monitor and startup there

Alternatives

aren't any?

Additional context

I did see another but that was because of wayland and was solved. this is on windows

Request Agreement

• [X] I have searched the existing issues and found no s...

[floral ledge]

@scenic hollow PR good to go

added a new readme change to mark off wayland support

[simple grove]

checkraisefold

[Vencord/Vesktop] New comment on pull request #326: feat: Add global shortcuts

Basic Windows and Wayland (via xdg-desktop-portal) support have been merged into Venbind! 🥳

There won't be another tagged version of Venbind until it is at least somewhat usable (confirmed with maintainer) - key release events aren't implemented and there's issues with the existing implementation that need to be fixed. Once that's done, and Darwin support is done (shouldn't be too hard - libuiohook has it inbuilt) then Venbind should be pretty ready, and then everything's on the JavaScrip...

[scenic hollow]

god does anyone here even have a macos machine to test it

[floral ledge]

that was my

FIRST thought

LMFAO

[scenic hollow]

honestly i think we can delay darwin support until someone comes around and contributes it i don't think it should block our release

[floral ledge]

like. we can see if it builds using a macos target because rust is Awesome but like..

A fair amount of people seem to use vesktop on macos

just from a basic search of this discord

[floral ledge]

@scenic hollow swapping over to napi-rs gives a lot more architectural freedom

how do you want this designed lol

separate thread for the platform-specific handler and the JS callback loop?

so 2 threads?

n-api has this delusional insane support under the threadsafefunction api that makes that possible

[scenic hollow]

yeah. we could theoretically try fixing up async for the libuiohook dispatch function so it's one thread but that might prove difficult. two threads it is

[scenic hollow]

i think that was actually why it was such a pain in the ass to implement it using napi rs
like for node bindgen you can literally have js callbacks with the same syntax as rust closures

[floral ledge]

Yeah node bindgen does it all for you

The napi-rs implementation is, indeed, a pain in the ass

It's much more flexible but its a huge pain

[scenic hollow]

btw i'm making progress for making windows work with the current ci cd

turns out libuiohook is very insistent on targeting linux despite having set the target to windows

[floral ledge]

lmao

[floral ledge]

Just realized uiohook isnt grabbing my mouse button events for my usual push to talk key

on windows

mouse4 is not a key I guess

separate event

EVENT_MOUSE_PRESSED

tragic

[scenic hollow]

is that what you see in the vesktop keybinds screen??

[floral ledge]

It's what I'm seeing looking at the libuiohook codebase

[scenic hollow]

huh

[floral ledge]

We'll end up needing to support that too

annoying as fuck

I imagine same issue on x11

[scenic hollow]

don't have mouse buttons to test

i mean i guess i could look at the source

[floral ledge]

hilariously I made some random internal napi function assertion fail immediately upon testing this new napi-rs code

its going well

[simple grove]

alany08

[Vencord/Vesktop] New comment on issue #375: Vesktop won't maintain Discord account logged in, need to relogin almost every time I open the app.

same here, been happening since starting to use vesktop. i have the NoDevtoolsWarning plugin and Developer Mode on, but the active account always gets logged out upon resume from suspend

[simple grove]

layershock

[Vencord/Vesktop] New comment on issue #628: [Bug] Flatpak Notifications fail to open chat when clicked on

I've noticed that it works just fine when running through XWayland

[humble mortar]

macos vm:

[simple grove]

Covkie

[Vencord/Vesktop] New comment on issue #1047: [Feature Request] Remember last position on startup (Windows)

can you show to contents of state.json at %AppData%\vesktop\state.json? Does the windowBounds array and displayid field change when closing in different positions?

[brazen kite]

I have a mac

I can test it

does the PR to vesktop need to be updated with the newer version? 🤔

actually I'll figure it out

[humble mortar]

mac isnt implemented at all

and yes the pr would need to be updated

[simple grove]

BlackFuffey

[Vencord/Vesktop] New comment on issue #372: A Variety of Screensharing Issues

I have a similar issue.

• If I share the whole screen, the stream will never load for other people.
• If I share a windowed application, the stream works. However it freezes the moment I fullscreen the application, and turns back normal if I window it again.
• If I share a fullscreened application, the stream does not load. If I window the application, stream will load instantly.
• If I share region of screen, the stream never loads, and for some reason the preview on host pc is showin...

[vernal lintel]

vesktop is cancelled

too many yappers saying the same thing 100 people have already said before them

[floral ledge]

finally

[hot juniper]

can't wait for vesktop 2

[vernal lintel]

people think if they repeat the same thing a 100th time it will get fixed quicker or smth

[floral ledge]

PICTURED: Vesktop developers working on global shortcuts

https://tenor.com/npdWkO1zYSW.gif

[simple grove]

littlecube23

[Vencord/venmic] New comment on issue #34: No sound gets shared in screenshare

Added a video showing the entire process as well as attaching a log now that I figured out how to do so. I also tried this time using only my headset mic and disconnected my blue yeti. The same thing occurs with both microphone workaround and without. Below is the video

Screencast from 2025-01-13 16-57-52.webm

Here is the log
venmic.log

...

[humble mortar]

i literally cannot repro this locally (streaming to myself with an alt) no matter the env but when i stream to a friend they get the cooked loading

I can consistently repro this with them in chromium by disabling and enabling hardware encoding

im confused why my local macos/windows vm or my phone doesnt have the issue but guh

webrtc routing shenanigans?

this is entirely a discord issue though

someonej should open a discord ticket

[floral ledge]

and you're trying on xdph?

Holy hell 1280 lines of code in xdph for screencopy

"disabling and enabling hardware encoding" is crazy

[floral ledge]

It could theoretically be a vesktop issue

Okay nevermind it uses electron's api which uses webrtc::DesktopCapturer

Okay the portal does matter webrtc uses xdp

wasnt too sure if they did before

[simple grove]

wiki-Bird

[Vencord/Vesktop] New comment on issue #375: Vesktop won't maintain Discord account logged in, need to relogin almost every time I open the app.

same here, been happening since starting to use vesktop. i have the NoDevtoolsWarning plugin and Developer Mode on, but the active account always gets logged out upon resume from suspend

does this still happen when you turn developer mode off?

[floral ledge]

@scenic hollow

How is the thing behind so far

Stable rustc is at 1.83.0

Thanks nix

Also why is mold disabled

[floral ledge]

The lockfile needs to be updated

idk how to do allat im on windows

with WSL of course

there

relatively easy with wsl

[scenic hollow]

yup everything is kind of out of date rn on main i was updating

[floral ledge]

dont you dare

i already updated the devenv lock

[floral ledge]

napi-rs cli is my personal 9/11

i hate github actions

i THINK it's gonna work this time

[scenic hollow]

i think github actions is hung on uiohook rn

[floral ledge]

had 2 change the target in devenv

which i missed

I hate nix

Bad software

@scenic hollow Ironically I got the CI working fully except uiohook-sys is now failing because cmake cant find ninja

Despite the fact that I have added it tot he devenv

I lied

it found it

This sucks

[scenic hollow]

yeah i've also had to manually specify where advapi32 is

and the windows.h include directory

[humble mortar]

the portal isnt the issue

well its not the issue for the non loading issue

idk about the artifacts

the hardware encoded enabled vs disabled, loading vs not thing can be replicated in pure chromium with vaapi so its 2000% not us

what's driving me insane though is i can't repro it purely locally

it always loads for me from my alt

I think I'll report it to discord and see what they say

[floral ledge]

can be replicated in pure chromium? LMFAO

[floral ledge]

figured that out immediately btw

you use a like

2 year old uiohook for some reason

way earlier than the latest commit at the time u started this repo so

newest one has a fixed cmake lists

[scenic hollow]

oh huh yeah i think there was issues with the latest commit at the time i don't remember what

[floral ledge]

Yeah there's some unknown type issues

I don't know how theres no issue for it

[simple grove]

Octogonee

[Vencord/Vesktop] Issue opened: #1048 [Bug] Screenshare working, but bugging out.

Discord Account

octogonee

Operating System

EndeavorOS

Linux Only ~ Desktop Environment

Hyprland

Package Type

AUR

What happens when the bug or crash occurs?

Screensharing is just weird.
Sometimes it doesnt load, and other times it does what is displayed in the images below.

I've been looking everywhere for a solution and found nothing.
Thanks!

What is the expected be...

[simple grove]

Psychloor

[Vencord/Vesktop] New comment on issue #1047: [Feature Request] Remember last position on startup (Windows)

{
    "firstLaunch": false,
    "windowBounds": {
        "x": 3870,
        "y": 139,
        "width": 1872,
        "height": 1161
    },
    "displayid": 274386696,
    "maximized": false,
    "minimized": true
}```

and now for some reason it works, weird

[floral ledge]

I spent forever looking for a fix only to read the error more closely and discover it was bindgen being stupid

.header("stdint.h")

World leading fix for mr worldwide

It is because uiohook is bad and uses C coder mind rot

Where the only reason it compiles is because they only include the specific header after stdint is included

However... EVIL binden with clang would not know about this since its trying to bindgen a single header file...

VICTORY IS MINE

https://github.com/checkraisefold/venbind/actions/runs/12765022276/job/35578352343

@scenic hollow

https://github.com/tuxinal/venbind/pull/3

turns out the build scripts were bugged before. you can't use cfg(target_os = abc) in a build script

have to use the cargo env variables

otherwise cross compilation explodes

[scenic hollow]

yeah i also noticed that turns out build.rs is different

so i was working on my own ci cd fixes and i was stuck at a part where you need some libpthread thing for compiling to windows and i'm wondering if all that could've been avoided by using x86_64-pc-windows-msvc instead

[floral ledge]

You have to use cargo-xwin though

napi-rs requires it for its cross builds and it solves 99% of issues so

i dont mind it

x86_64-pc-windows-gnu is a bad stupid target

dont use it

[floral ledge]

cargo-xwin is fascinatingly peak tbh. Generates a cmake toolchain for you for any build scripts using cmake-rs and automatically injects it into the environment with no intervention needed, automatically sets up clang-cl and llvm-lib for MSVC abi compat

Pulls in windows-msvc-sysroot to use the windows header set

[scenic hollow]

yeah it's so cool dunno why i haven't heard of it

[floral ledge]

only 376 github stars is crazy

[floral ledge]

The cools just use this nowadays i think https://github.com/cross-rs/cross

Free of charge multi platform building and testing i guess

Nevermind it sucks it just spins up a bunch of docker containers

:(

So boring!!

That doesn't count as cross compilation thats just docker abuse

[simple grove]

SebSK3

[Vencord/Vesktop] New comment on issue #1048: [Bug] Screenshare working, but bugging out.

Probably related: #1012 and #372

[simple grove]

Covkie

[Vencord/Vesktop] Issue closed: #1047 [Feature Request] Remember last position on startup (Windows)

Covkie

[Vencord/Vesktop] Issue closed: #1048 [Bug] Screenshare working, but bugging out.

[humble mortar]

what if chromium wasnt bad and everything just worked

[simple grove]

Octogonee

[Vencord/Vesktop] New comment on issue #1048: [Bug] Screenshare working, but bugging out.

I really dont think this is a duplicate.
I'm using an AMD gpu, and no issue has this exact problem.

[simple grove]

Luke-Sims

[Vencord/Vesktop] Issue opened: #1049 [Bug] Can't access to Documents and Music directory when wanting to upload a file

Discord Account

No response

Operating System

Debian 12 bookworm

Linux Only ~ Desktop Environment

Gnome on Wayland

Package Type

stable 358711 (f042b2b) Build Override: N/A Vencord 3243120 (Vesktop v1.5.4) Electron 33.2.1 Chromium 130.0.6723.137

What happens when the bug or crash occurs?

When I want to upload a file

What is the expected behaviour?

I can't access to files in Documents and Music, It appear one day (I had access before)
An update ? I have ac...

[simple grove]

Vendicated

[Vencord/Vesktop] Issue closed: #1049 [Bug] Can't access to Documents and Music directory when wanting to upload a file

Vendicated

[Vencord/Vesktop] New comment on issue #1049: [Bug] Can't access to Documents and Music directory when wanting to upload a file

if you're using flatpak - either don't, or grant vesktop access to the desired folders with flatseal

otherwise, electron/chromium/system issue so nothing we can do, it will likely fix itself in a future update of any of those

[humble mortar]

probably the electron changing to a new portal version before it was stable thing

based on the screenshots those pickers are different

so likely their de doesnt sync the pinned stuff between portal impls

speaking of-

34.0.0 hit stable

which has the fix for that

soo

[simple grove]

Richy-Z

[Vencord/Vesktop] Issue opened: #1050 [Feature Request] Add menu bar icon for Vesktop on macOS

Discord Account

iirzd

Motivation

It is often quite nice to have quick control over your input and output, especially when in a hurry. Discord has a menubar icon on macOS similar to Windows, allowing you to mute and/or undeafen:

Vesktop however, lacks this important feature.

Solution

Implement a menu bar icon into vesktop for macOS. If it isn't present on...

[hot juniper]

what kind of alternate universe do you wish you were living in

[floral ledge]

Replace chromium with firefox and make gecko great

Peak alt universe...

[simple grove]

gmanskibiditoilet

[Vencord/Vesktop] Pull request opened: #1051 Adding support for 8K and 120FPS to screenshare

Add further support for 8K streaming as well as an option to stream 120 FPS as well.

A modified version of the original pull request #1024 .

gmanskibiditoilet

[Vencord/Vesktop] New comment on pull request #1024: Add 4k resolution to screenshare

great job bruh, we have just added another pull request which further allows the user to choose 8k and/or 120 FPS screenshare #1051

gmanskibiditoilet

[Vencord/Vesktop] New comment on pull request #1024: Add 4k resolution to screenshare

hi @Vendicated can you please help review and merge this pull request? TYSM

[hot juniper]

hu

don't these crash clients

just add source option

[heady merlin]

great job bruh, we have just added another pull request which further allows the user to choose 8k and/or 120 FPS screenshare #1051

what is this comment lol

[hot juniper]

yapping

i doubt anyone is seriously using discord to stream in 8k resolution

[gentle hatch]

isn't there a bitrate cap anyways, so it would look horrible

[vernal lintel]

yes

that's insane

[simple grove]

Vendicated

[Vencord/Vesktop] Pull request closed: #1024 Add 4k resolution to screenshare

Vendicated

[Vesktop:main] 1 new commit

5a72491 Add 4k resolution to screenshare (#1024) - makindotcc

Vendicated

[Vencord/Vesktop] Pull request closed: #1051 Adding support for 8K and 120FPS to screenshare

[rotund panther]

chrrybmb64

[Vencord/PKGBUILDs] Pull request opened: #1 [vesktop-electron'

[rotund panther]

Covkie

[Vencord/PKGBUILDs] New comment on pull request #1: [vesktop-electron] use electron33 instead of electron

the point of the -electron packages are to use the system/latest electron. If you wanted to use a supported electron version you should use simply use vesktop

[vernal lintel]

hows electron 34

good bad buggy?

[humble mortar]

seems fine

I havent had a chance to check streaming with it yet tho

it just had its first stable release like yesterday tho so maybe more niche bugs will popup

[rotund panther]

chrrybmb64

[Vencord/PKGBUILDs] New comment on pull request #1: [vesktop-electron] use electron33 instead of electron

There is no benefit to using the latest electron, it makes no difference other than increasing the possibility of something breaking. Using electron33 gives all the benefits of system electron but without that downside. It's the best of both worlds.

[vernal lintel]

there really isnt any benefit to using system electron in the first place

the only benefit is more up to date electron

[humble mortar]

EXACTLY 😭

[rotund panther]

Covkie

[Vencord/PKGBUILDs] Pull request closed: #1 [vesktop-electron] use electron33 instead of electron

Covkie

[Vencord/PKGBUILDs] New comment on pull request #1: [vesktop-electron] use electron33 instead of electron

If you're worried about something breaking, don't use system electron and use what is bundled & supported. The system electron packages I'm aware of use electron as that is their purpose — to use the latest electron. Pinning a specific version will be no different than using bundled.

[simple grove]

2meito

[Vencord/Vesktop] Issue opened: #1052 [Feature Request] Screensharing without capturing window decorations

Discord Account

No response

Motivation

Screensharing windowed applications also shares their window decorations (e.g. titlebar).

Solution

Screensharing should not include window decorations similar to the behavior of screensharing windowed applications on Windows.

Alternatives

Running applications at full screen or without window decorations is possible, but not ideal for all applications.

Additional context

No response

Request Agreement

• [X] I have s...

Covkie

[Vencord/Vesktop] Issue closed: #1052 [Feature Request] Screensharing without capturing window decorations

Covkie

[Vencord/Vesktop] New comment on issue #1052: [Feature Request] Screensharing without capturing window decorations

You LOVE client side decorations (we have no control over this sorry :/)

[humble mortar]

@vernal lintel where did you see discord had the detectable.json deltas?

[simple grove]

pendo324

[Vencord/Vesktop] Pull request opened: #1053 feat: move arrpc server into a worker thread

Solution for #1009, which I can reliably reproduce without this change.

I have another version of this which instead works with this arrpc PR https://github.com/OpenAsar/arrpc/pull/133, but it seems like Vesktop might be the appropriate place to put this complexity.

I have tested that game status RPC works as expected but I haven't found a game with which to test invites.

Covkie

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

invite events are simply opening https://discord.gg/D9uwnFnqmd in your browser of choice and it sending that to your open client. The callback is so the website can display its completed or error thing

Same case as https://github.com/Vencord/Vesktop/pull/1016

[rotund panther]

chrrybmb64

[Vencord/PKGBUILDs] New comment on pull request #1: [vesktop-electron] use electron33 instead of electron

If you're worried about something breaking, don't use system electron and use what is bundled & supported

System electron is just as stable as bundled electron if you're using the right version

The system electron packages I'm aware of use electron as that is their purpose — to use the latest electron

Not a single electron app in arch repositories uses the electron package. They all use specific versions. Go and look. The same also applies for the majority of electron apps in t...

[vernal lintel]

its in the client

open network tab while launching

[simple grove]

Covkie

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

anyway yeah doesn't work for invites

[vernal lintel]

why did they add some dependency 😭

[humble mortar]

why are they hardcoding the types like that

crazy

i was crazy once

[humble mortar]

it pulls the whole thing

every time

[vernal lintel]

oh yeah deltas is the banned domain hashes

[humble mortar]

EXPLODE

https://github.com/OpenAsar/arrpc/issues/134

holy moly

wonder if they'll do it in a native language

[vernal lintel]

they dmed me about this earlier

[humble mortar]

will it still be js?

[vernal lintel]

idk

ask ducko

maybe ts since they seem to like it now

[vernal lintel]

(we're discussing rn, they are open to native or js)

[simple grove]

pendo324

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

invite events are simply opening https://discord.gg/D9uwnFnqmd in your browser of choice and it sending that to your open client. The callback is so the website can display its completed or error thing

Same case as #1016

Oh, I thought it was game invites, lol. Thanks for the info. Let me see what's up. It should be fixable, I think

pendo324

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

Off by one error, works on my side now

[vernal lintel]

sorry i mixed them up

epic arrpc rewrite thread

i dont remember who else was interested in arrpc stuffs but ^ @humble mortar

[simple grove]

Covkie

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

Instead of writing an insane review with github web editor heres a patch. It's functionally the same but doesn't have the hardcoded types (?) or add a new dep

buh.patch

[simple grove]

pendo324

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

Instead of writing an insane review with github web editor heres a patch. It's functionally the same but doesn't have the hardcoded types (?) or add a new dep

buh.patch

Thanks. I removed the new dependency in my latest revision. Not sure why we wouldn't want the types?

[simple grove]

slepice24

[Vencord/Vesktop] New comment on issue #919: [Bug] Vencord randomly restarts

UPDATE - I have upgraded form Nvidia to AMD
same bug with AMD gpu most likely something with Wayland? but didnt tested it on Xorg
It happens a lot when playing game on second virtual screen

[simple grove]

AlexDiego123

[Vencord/Vesktop] Issue opened: #1054 [Bug] Custom title bar not working on Windows

Discord Account

@alexdiego123

Operating System

Windows 10

Linux Only ~ Desktop Environment

No response

Package Type

Setup .exe for Windows

What happens when the bug or crash occurs?

Vesktop doesn't show Discord's custom title bar even tho i enabled the setting.

What is the expected behaviour?

Vesktop should show Discord's custom title bar instead of the Windows title...

[simple grove]

Covkie

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

You're over complicating it by quite a bit, sending the data over postMessage is fine, you don't need to handle the port stuff yourself & worker.on error exists so you dont need a try catch.
patch.txt

Also please stop forcing pushing it makes it insane to follow changes :sob:

[woeful current]

@humble mortar i force push all prs you ever view

[humble mortar]

ill eat you alive

[simple grove]

ndelta0

[Vencord/Vesktop] Issue opened: #1055 [Bug] White screen on system startup without internet

Discord Account

madpie._

Operating System

Windows 11

Linux Only ~ Desktop Environment

No response

Package Type

Setup EXE

What happens when the bug or crash occurs?

The same symptoms as #969. It only happens when the system starts from a cold boot (i.e. from shut down or restart) when there's no internet connection available yet.

I have created a new issue to mention that this also happens on Windows (the other issue is locked down to collaborators).

What i...

[hot juniper]

this is a real issue

hu but also understandable but also not

[simple grove]

pendo324

[Vencord/Vesktop] New comment on pull request #1053: feat: move arrpc server into a worker thread

You're over complicating it by quite a bit

Agree to disagree, I suppose. It's much more readable to me with the MessageChannel, and adds like... 1 line of code, which is why I didn't use postMessage. I’m surprised your code works using a .ts file for the worker, even when it isn’t explicitly loaded using an import/require (you reference the js output, which I’m surprised is included by ESBuild in the bundle when it’s not ever imported). Let me try removing the new entrypoint I added ...

[simple grove]

103sbavert

[Vencord/Vesktop] New comment on issue #753: [Bug] Dismissing screenshare request causes crash

Is there an Electron repo issue that is opened regarding the regression with electron? I can't seem to find it. If there's one, I'd like to read the discussion.

[hot juniper]

i found this in my emails before clicking into this channel

[grand swift]

@vernal lintel you

apparently gamescope exposes a pipewire stream named gamescope

for video

u don't need the special protocol

is there a way we can force electron to use pipewire capture on x11

[hot juniper]

you probably know a lot more than I do about this but why does vesktop need to run in x11 with a wayland compositor

[grand swift]

gamescope is a Wayland compositor x11 server

all clients use x11

[hot juniper]

wtf is valve cooking

[grand swift]

if you set a flag it will expose Wayland instead but it's WIP and I tried vesktop with it and it segfaults

[grand swift]

most games are x11 and steam is x11

[hot juniper]

i mean they probably think chromium on wayland is horror (it still is)

[grand swift]

But we need pipewire capture to capture properly on deck game mode

Not x11 capture

X11 works but only for some games

[hot juniper]

Sounds like you might need to patch chromium hu

[grand swift]

pipewire will work for everything because it's gamescope's internal video streams

https://github.com/chromium/chromium/blob/631150176d83afa7c76c6403e959b4a1e87ae0f7/content/public/browser/desktop_capture.cc#L119

[sour mural]

desktop_capture.cc: Line 119

return session_type == base::nix::SessionType::kWayland;

[grand swift]

I HATE GOOGLE

they enable pipewire if ur using a Wayland session explicitly

[grand swift]

ok in theory

setting xdg session type Wayland and Wayland display + ozone platform x11 should satisfy the conditions for pipewire desktop capture in chromium and webrtc

[hot juniper]

try it?

[grand swift]

not home

just saying it so I don't forget

I imagine this might have issues because it doesn't support the capture picker portal

And pipewire in webrtc probably expects that

[simple grove]

compiler-warning

[Vencord/Vesktop] New comment on issue #1038: [Bug] closing and reopening vesktop does not work, claims its already running

think it's safe to say at this point that doing a complete reinstall seems to have fixed the issue

[grand swift]

@hot juniper i tested

hard crash when i hit screenshare

xddddd

[hot juniper]

[grand swift]

we need to make fake xdg desktop portal wrapper or some shit

or figure ou ta way to make it not prompt

[humble mortar]

gamescope doesnt have a screenshare portal

youre fighting a useless battle

[brazen kite]

it might have a wayland protocol for keybinds

[brazen kite]

update: I don't think it does

[floral ledge]

@scenic hollow good changes in latest commit

thanks for the cargo-xwin cahnge to nixpkg

i am intrigued if the windows bindings still build with that change

if they do thats wonderful

also we may not need the napi cli - the cli cross compilation just wraps cargo-xwin for Windows

and cargo-zigbuild for other platforms

since we ignore most of the cli output anyways

also the type bindings right now are technically wrong - they still work but the file just exports functions instead of a class

napi can also generate a full .js file wrapper that does all the platform-specific module name loading/finding for you but that wouldn't be consistent with vesktop's manual thing used for venmic/is heavier

[simple grove]

danielfrrrr

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

Although the VaapiVideoDecoder and VaapiVideoDecodeLinuxGL options are enabled by default, they have also been renamed: AcceleratedVideoDecoder and AcceleratedVideoDecodeLinuxGL.

See: https://chromium.googlesource.com/chromium/src/+/refs/heads/main/docs/gpu/vaapi.md#vaapi-on-linux

[humble mortar]

when im in a repeating known information challenge and my opponent is danialfrrrr

[simple grove]

Covkie

[Vencord/Vesktop] New comment on issue #753: [Bug] Dismissing screenshare request causes crash

@103sbavert I finally got around to reporting it 2 days before your comment :p https://github.com/electron/electron/issues/45198

[simple grove]

103sbavert

[Vencord/Vesktop] New comment on issue #753: [Bug] Dismissing screenshare request causes crash

By the way, to the desktop team, you guys could easily fix this issue with the latest discord 0.80

[scenic hollow]

they did as far as my testing went!

[simple grove]

Sid127

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

Electron 34 is out with the chromium version that has these new flags

[scenic hollow]

also wdym by this? venbind.d.ts exports a class containing functions. what's it supposed to be?

[floral ledge]

the new napi-rs bindings is just require() and directly accessible functions not wrapped in a class, but the .d.ts still seems to work correctly

not sure if a class is proper though

since the napi-rs bindings dont export a class

they just export functions

if you build with the napi-rs cli you'll see what I mean

look at the .d.ts file it generates

[scenic hollow]

omg i read that backwards again i thought you meant it's supposed to be a class but it's functions guhh

[floral ledge]

node bindgen exports a class i think

[scenic hollow]

okok i'll try

[floral ledge]

napi-rs just exports functions

classes require extra work

[simple grove]

RSKYS

[Vencord/Vesktop] Pull request opened: #1056 README.md: Updated info.

I thought it's not bad people to add extra info, I recently came around this and it made wayland screensharing less pain in the neck,

Thank you guys for developing this!

Vendicated

[Vesktop:main] 1 new commit

7a19c81 README.md: add slackware link to unofficial rep... - RSKYS

Vendicated

[Vencord/Vesktop] Pull request closed: #1056 README.md: add slackware link to unofficial repos

[simple grove]

ninja-

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

I would appreciate fixing AMD VA-API encoding even before Electron 34 is used, so adding PlatformHEVCDecoderSupport,VaapiVideoDecoder,VaapiVideoEncoder,VaapiIgnoreDriverChecks, there is no way to override these flags without building whole Vesktop manually

Vendicated

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

there is no way to override these flags without building whole Vesktop manually

this is just not true. you can pass them via command line flag vesktop --enable-features=Feat1,Feat2,Feat3

ninja-

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

ok, I missed the part where Vesktop reads passed enable-features and merges them now I see it. without it, Chrome doesn't merge enable-featuers and would pick the first one or last one or something

ninja-

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

Unfortunately, this segfaults with no stacktrace or useful logs after which Vencord restarts with disabled hardware acceleration. I'll try again later with Electron 34.

flatpak run dev.vencord.Vesktop --enable-features=VaapiVideoEncoder,VaapiIgnoreDriverChecks,VaapiVideoDecoder,PlatformHEVCDecoderSupport

[vernal lintel]

i think if you enable VaapiIgnoreDriverChecks you probably should be expecting potential crashes

live footage of electron devs releasing new electron version https://tenor.com/view/bugs-video-games-software-developer-gif-25721930

[simple grove]

CelDaemon

[Vencord/Vesktop] New comment on issue #372: A Variety of Screensharing Issues

I'm currently having an issue with Vesktop sometimes completely freezing after clicking the button to start a screenshare, then selecting my main monitor in the KDE dialog.

It does open the menu to select the audio channel and other settings, but the application completely freezes after that, including the system tray menu. It's still possible to talk in the VC and hear things however.

I'm having this issue on Fedora 41 KDE, using Wayland and the official .rpm package

ninja-

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

after bumping electron 34 it seems to work correctly without segfault with this set of flags:

diff --git a/src/main/index.ts b/src/main/index.ts
index 4eb863d..2c04a98 100644
--- a/src/main/index.ts
+++ b/src/main/index.ts
@@ -35,7 +35,7 @@ function init() {
     if (hardwareAcceleration === false) {
         app.disableHardwareAcceleration();
     } else {
-        enabledFeatures.push("VaapiVideoDecodeLinuxGL", "VaapiVideoEncoder", "VaapiVideoDecoder");
+        enabledFeat...

Vendicated

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

We definitely should not add likely unstable flags like VaapiIgnoreDriverChecks. Users should experiment with such flags themselves if needed

ninja-

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

This flag is nothing else than a whitelist of vendors being just "Intel". This combination of flags is known to be fully stable in recent Chromium versions with AMD.
You can't get around Google's stubbornness in heavy feature gating of this stuff and they don't plan on changing policy any time soon no matter how stable VA-API is and no matter how much testing it gets.

https://source.chromium.org/chromium/chromium/src/+/main:media/mojo/services/gpu_mojo_media_client_linux.cc;l=144?q=Vaap...

Vendicated

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

that piece of code explicitly states it is broken on Novideo though. Does that mean it will possibly crash on NVIDIA GPUs? If yes, that would be unacceptable. Perhaps it should only be enabled if not using Novideo

ninja-

[Vencord/Vesktop] New comment on pull request #1007: Change `VaapiVideoEncoder` to `AcceleratedVideoEncoder`

This does nothing on Nvidia because Nvidia does not ship any VA-API backends.

There's a community driver but it doesnt support encoding so not useful for streaming https://github.com/elFarto/nvidia-vaapi-driver

[simple grove]

martyfuhry

[Vencord/Vesktop] New comment on issue #375: Vesktop won't maintain Discord account logged in, need to relogin almost every time I open the app.

This has been happening to me all of a sudden, as well. I do not have Developer Mode on and do not have any plugins. I am running the Flatpak on Pop!_OS with the new Cosmic DE.

[simple grove]

fnogcps

[Vencord/venmic] New star added

[scenic hollow]

does eslint-disable-next-line no-useless-escape not do anything anymore

nvm i thought it would give an error when running