f_ changed the topic of #wayback to: Wayback - a wayland-based X11 environment | https://wayback.freedesktop.org/ | src: https://gitlab.freedesktop.org/wayback/wayback | logs: https://libera.catirclogs.org/wayback | matrix bridge: #wayback:catircservices.org
<Consolatis[m]1> maybe the whole thing could be turned around. instead of adding custom wayland protocols and teach xwayland to use them one could also make the compositor open a unmapped X11 window and then listen to property changes and/or client messages there
grufwub has quit [Remote host closed the connection]
grufwub has joined #wayback
wezm has quit [Remote host closed the connection]
cow has quit [Remote host closed the connection]
<whitequark[cis]> does it matter what the upstream wayland security philosophy is?
<whitequark[cis]> whatever it is, they're clearly not very good at making it, or making it happen. and a private extension doesn't have to be approved by anyone
<ConanKudo[m]1> only if we want to have something standardized or need to adapt xwayland
<whitequark[cis]> right
_whitelogger has joined #wayback
jvvv has quit [Quit: WeeChat 4.6.3]
ThinkT510 has quit [Quit: ThinkT510]
__0x1eaf has joined #wayback
weirdtreething has quit [Remote host closed the connection]
weirdtreething has joined #wayback
jvvv has joined #wayback
<dramforever[m]1> why is there a SIGSEGV handler in xwayback and xwayback only
<dramforever[m]1> unix is a mistake
<f_> 23:03 <Wayback> @Conan_Kudo closed issue: #44 Criteria for an initial release (https://gitlab.freedesktop.org/wayback/wayback/-/work_items/44) 23:05 <ConanKudo[m]1> I have created the %"0.1" milestone 23:05 <ConanKudo[m]1> https://gitlab.freedesktop.org/wayback/wayback/-/milestones/1#tab-issues
<Wayback> #44 Criteria for an initial release (https://gitlab.freedesktop.org/wayback/wayback/-/work_items/44) [closed]
<f_> Nice
<f_> (Also yes I also believe dbus doesn't have its place in Xwayback)
<f_> navi, !70 I just noticed you remove -terminate
<Wayback> !70 xwayback: simplify argument parsing and main body (https://gitlab.freedesktop.org/wayback/wayback/-/merge_requests/70) [opened]
<f_> Can you add it back?
norwoodites has joined #wayback
pinskia has quit [Ping timeout: 276 seconds]
<f_> ConanKudo[m]1: hi, can you add #50 to %0.1 milestone?
<Wayback> #50 protocol error when switching terminal (https://gitlab.freedesktop.org/wayback/wayback/-/issues/50) [opened]
<f_> I think that's something we need to handle properly for an initial nonstable release
<f_> (thanks dok!)
<dok> np :)
<f_> navi: and tbh, I was going to improve args handling anyway, just didn't have much time in the past few days.
<f_> I had some ideas for better opt handling
<f_> but I can wait
pinskia has joined #wayback
norwoodites has quit [Ping timeout: 240 seconds]
<navi> f_: i'll fix that in a moment, and also see if moving to posix_spawn works
<dramforever[m]1> nice
<ConanKudo[m]1> f_: done
<navi> do we want to list all Xwayland+Xserver+Xorg flags in -help?
<f_> navi, ConanKudo[m]1: thanks!
cow_ has joined #wayback
<f_> I'm not sure what advantages does posix_spawn have over fork+exec
<navi> it's more efficient (fork is really slow, *even* with CoW memory pages) and it's not gonna have signal timing issues
<navi> (with fork+exec, we need to mask sigchld around the fork call for it to not have issues)
<navi> i did have an issue on openrc exactly bc of that, `pid = fork()` but the child would actually die before `pid` got assigned in the parent
<navi> so the signal handler ran, and `pid` was -1, so it didn't reap
<navi> (or rather, it did, but the event loop later on didn't get the message it wanted
<navi> )
Guest62 has joined #wayback
Guest62 has quit [Client Quit]
<f_> navi: ok fair enough
<f_> navi: submitted some more review comments to your MR
<f_> but overall this is pretty nice, thank you!
<f_> navi: And by the way I do not currently deem necessary to copy Xserver args *now*, my comment about -help was just me being curious
<f_> I can do it myself
cow_ has quit [Quit: Konversation terminated!]
<navi> i can just revert back to calling a function
<navi> printing `Xwayback [options ...]`
<navi> and we add the full list later
<f_> yep
<navi> i was kind of avoiding it bc i really didn't want to manually insert all X flags in there and such
<navi> but idk can't think of any alternatives
<f_> like I said you don't need to
<f_> I've some ideas for how best to do it
<f_> (I mean you don't need to do it in your MR)
<navi> yeah i got it
cow has joined #wayback
Pixi` has joined #wayback
Pixi has quit [Ping timeout: 260 seconds]
Pixi` is now known as Pixi
joast has quit [Ping timeout: 248 seconds]
<navi> latest version of the mr broke things somehow
joast has joined #wayback
<navi> > wayback-compositor: render/drm_syncobj.c:244: wlr_drm_syncobj_timeline_waiter_finish: Assertion `wl_list_empty(&waiter->events.ready.listener_list)' failed.
<navi> huh
caskd has quit [Ping timeout: 272 seconds]
arraybolt3 has quit [Quit: Konversation terminated!]
caskd has joined #wayback
orowith2os[m] has quit [Quit: Idle timeout reached: 172800s]
<f_> navi: hmmm you didnt touch the compositor right?
<navi> nope
<f_> Then I wouldn't worry too much about that
<f_> Maybe add a comment to the MR about it
<f_> But I suspect it's another dormant bug similar to the ctrl-alt-BackSpace segfault
<f_> I'll do one more review to your MR later
<f_> But I think it should be all good ^^
<navi> okay my desktop it works
<navi> my laptop can reliably hit that crash
<navi> i wonder if making it faster (by not forking) is what allows the bug to show
<dramforever[m]1> posix_spawn is definitely forking on linux for some definition of forking
<navi> it's not a fork, it's clone(CLONE_VFORK)
<navi> which leaves the memory map / pages untouched
<dramforever[m]1> yeah that could change the timing
<f_> navi: in any case I wouldn't worry too much for now
<f_> navi: ok, this LGTM now
<f_> Thank you!
<f_> axtlos, ConanKudo[m]1 - do you have any more comments for !70 ?
<Wayback> !70 xwayback: simplify argument parsing and main body (https://gitlab.freedesktop.org/wayback/wayback/-/merge_requests/70) [opened]
<axtlos> skimming over it, looks good
<axtlos> i trust navi to not do anything bad
<f_> oh I definitely don't trust navi to not put a rootkit in wayback
<f_> hehe
<f_> anyway if there's no more comments I believe it's good to merge
<f_> the args stuff I think I'll see if I can rework it again later
<navi> (why was my VLA question thread auto-resolved)
<navi> (gitlab???)
<f_> oh it's just gitlab being very smart as usual
<navi> i don't know what criteria it has to "resolve" a thread automatically on push
<f_> it autoresolves if the line referenced changed
<f_> assuming the line has been fixed
<f_> TBH it would be really awesome if someone went to pick the best of sourcehut and the best of gitlab and merge them together
<f_> would be interesting to see the result
<navi> i want to do that
<f_> if only we all had infinite time, right?
<navi> yeah
<dramforever[m]1> in case anyone was wondering about me my troubles with trying to run wayback on my other laptop, it's notourbug, i can repro it on "normal" xwayland, gonna start knocking on other doors starting from mesa (which is the thing that's different) https://gitlab.freedesktop.org/mesa/mesa/-/issues/13564
<f_> both good and bad news
<ConanKudo[m]1> f_: I could turn off the autoresolve on diff change
<f_> ConanKudo[m]1: oh that'd be awesome
<ConanKudo[m]1> in general I usually don't because people forget to resolve discussions all the time
<f_> also would be nice to give the ability for people to resolve threads
<ConanKudo[m]1> but if we want it switched off, I can
<ConanKudo[m]1> f_: that is not possible
<ConanKudo[m]1> only project members and the MR creator can resolve threads
<f_> Oh, so anything higher than Guest I suppose? Or something
<ConanKudo[m]1> yeah
<ConanKudo[m]1> developer or higher I think
<dramforever[m]1> wait, does unresolved threads block merge now?
<f_> yes, they do
<ConanKudo[m]1> it always has
<dramforever[m]1> can we turn this off
<ConanKudo[m]1> it's a setting, but fdo gitlab defaults to it
<f_> why?
<f_> I find it useful
<f_> (threads blocking merge that is)
<ConanKudo[m]1> I would probably not disable autoresolve alongside maintaining thread blocking merge
<dramforever[m]1> yeah
<f_> hmm fair enough
<dramforever[m]1> that's what i was thinking as well
<f_> usually autoresolve kinda works I guess
<ConanKudo[m]1> yeah, it's rare that it doesn't
<ConanKudo[m]1> and tbh, a new thread should be made for new diffs anyway
<f_> ok, then I guess it's reasonable to keep it on
<ConanKudo[m]1> fun fact, long ago at $dayjob[-2], my team contributed that feature (toggling autoresolve)
<f_> cool fun fact :)
<ConanKudo[m]1> almost 8 years ago I think
<dramforever[m]1> i think autoresolve is fine
<f_> time flies fast
<dramforever[m]1> it's not perfecft
<ConanKudo[m]1> nothing really is
<dramforever[m]1> neither is my typing apparently
<ConanKudo[m]1> pobodys nerfect inded
<dramforever[m]1> but yeah it does its job, mostly, and it's easy to override anyway
<f_> neither is my fullscreen not actually fullscreen lxterminal
<dramforever[m]1> f_: what was the good and bad news again
<f_> dramforever[m]1: good news = not our bug, bad news = someone else's bug
<dramforever[m]1> ah, my news, haha
<f_> yes :)
<dramforever[m]1> thought you had news to report
<f_> nah I don't at the moment
<dramforever[m]1> bad news = it still just doesn't work
<dramforever[m]1> i was going to ask to block 0.1 if it's ourbug, because then we might have a rush in of users finding their kde crashes actually
<f_> oh 0.1 doesn't have to be perfect
<dramforever[m]1> well it still crashes, but we can blame someone else now
<ConanKudo[m]1> and kde isn't going to be blocking criteria
<f_> right now the main thing is getting wayback to a point where 'simple' environments works just fine
<f_> stuff like icewm, windowmaker, etc
<ConanKudo[m]1> and progman :)
<f_> fvwm3... and obviously also progman, yes :)
<dramforever[m]1> sure, but i was thinking like, well who knows what else might be broken
<dramforever[m]1> turns out, other software
<f_> Then we can progressively raise requirements
<f_> to bigger DEs
<dramforever[m]1> fwiw kde works completely fine on asahi linux for me
<dramforever[m]1> like it really just works
* ConanKudo[m]1 shrugs
<dramforever[m]1> i find this very impressive honestly
<f_> dramforever[m]1: you mean, KDE works in Wayback in Asahi? Or s/in Wayback// ?
<dramforever[m]1> in wayback, on asahi
<f_> oh that's cool
<dramforever[m]1> which has a different gpu and different driver, obviously
<dramforever[m]1> do images go through to irc? like as a link?
<ConanKudo[m]1> they show up as a link
<dramforever[m]1> yeah that's the entire kde x11
<dramforever[m]1> i'm not sure what i expected when i put startplasma-x11 in .xinitrc and did wayback-session, but it definitely wasn't "it just works, it turns out, especially if i use my docked setup with mouse and separate kb"
__0x1eaf has quit [Quit: Lost terminal]
<whitequark[cis]> nice
<f_> That's really cool
<f_> I remember whitequark brought up that it's not working
<whitequark[cis]> i believe the reason it seems to not work is that i have two monitors
<whitequark[cis]> it most likely has nothing to do with startplasma-x11 specifically
<f_> I guess it might be how it's setup there? Not to say you should change anything when it works on regular Xserver
<whitequark[cis]> i created a new user just to be sure i don't have some detail of my setup mess things up
<f_> whitequark[cis]: I mean, your setup, if it works in Xserver, should definitely work in wayback
<f_> wayback is supposed to be Xserver-compatible
<f_> If something works in Xserver and doesn't in wayback that's a bug
<whitequark[cis]> ah yes. but the multi-monitor issue is a known one
<f_> yes
<dramforever[m]1> you could set the monitor now
<dramforever[m]1> it's a hack but it ... kinda works to pick one monitor
<dramforever[m]1> <whitequark[cis]> i believe the reason it seems to not work is that i have two monitors
<dramforever[m]1> this was back before $WAYBACK_OUTPUT
<dramforever[m]1> in fact
<dramforever[m]1> you can say it's way back
<anabelae[m]> lol
<whitequark[cis]> hm
<whitequark[cis]> how do i use WAYBACK_OUTPUT?
<dramforever[m]1> grep connectorName ~/.config/kwinoutputconfig.json
<dramforever[m]1> you have stuff like HDMI-A-1 or eDP-1
<dramforever[m]1> that's what i have anyway, so you can do WAYBACK_OUTPUT=HDMI-A-1 wayback-session
<dramforever[m]1> ... as we've discussed extensively last time
<dramforever[m]1> the alternative is the monitor make and model with a space in between which is potentially more stable
<dramforever[m]1> (hope you don't have two identical monitors?)
<whitequark[cis]> hm.
<whitequark[cis]> ```
<whitequark[cis]> ```
<whitequark[cis]> "connectorName": "WL-0",
<whitequark[cis]> "connectorName": "X11-0",
<dramforever[m]1> hmm
<whitequark[cis]> oh wait... this requires me to run kwin on a normal x11 session
<whitequark[cis]> i, uh, don't use kwin :D
<dramforever[m]1> ... interesting
<dramforever[m]1> oh
<dramforever[m]1> yeah that explains it, i thought you tried kwin because you use kwin regularly
<whitequark[cis]> i use kde with i3
<dramforever[m]1> yeah you never said you use kwin did you...
<dramforever[m]1> oops
<dramforever[m]1> xrandr --listmonitors then?
<dramforever[m]1> that looks like this for me
<dramforever[m]1> ```
<dramforever[m]1> Monitors: 2
<dramforever[m]1> 0: +*HDMI-A-1 4800/632x2700/360+3025+0 HDMI-A-1
<dramforever[m]1> 1: +eDP-1 3024/302x1890/189+0+0 eDP-1
<dramforever[m]1> ```
<whitequark[cis]> ok so what happens is that plasmashell crashes
<whitequark[cis]> and i can't quite figure out why
<dramforever[m]1> oh no
<dramforever[m]1> do you have like coredumps
<whitequark[cis]> it dies in libgallium
<dramforever[m]1> iris_set_damage_region?
<whitequark[cis]> i'm going to find out after gdb downloads like ten gigabytes of debug inof
<whitequark[cis]> s/inof/info/
<dramforever[m]1> uh, you shouldn't need that i think
<dramforever[m]1> unless whatever distro it is you're using aggressively strips even the basic .symtab
<dramforever[m]1> qt debug symbols is like, one gig for each library
<dramforever[m]1> i actually don't know why the coredumpctl backtrace is so bad
<whitequark[cis]> this reminds me of how bad gdb is
<dramforever[m]1> it might print a backtrace
<whitequark[cis]> why does ksplashqml link to libusbmuxd
<dramforever[m]1> wait i think last time i checked gdb times out the download for the debuginfod thing at 90s by default? which is very ???
<dramforever[m]1> so by default if download takes 90s it goes "guess i'm never getting it"
<whitequark[cis]> yes, iris_set_damage_region
<dramforever[m]1> right, so xwayland + kwin + iris just diesn
<dramforever[m]1> *dies
<dramforever[m]1> same deal on my "other" laptop
<dramforever[m]1> same thing, and as said not wayback-specific
<whitequark[cis]> have you reported it
<whitequark[cis]> * reported it?
<dramforever[m]1> because that's where it crashes and iris vs asahi one of the two variables that changed (well, the other is x86_64 vs aarch64 but i have no other computers)
<dramforever[m]1> ugh
<dramforever[m]1> i wonder if they'll blame it on nvidia
<dramforever[m]1> i'm fairly sure this happens without nvidia, it's not even remotely involved
<f_> whitequark[cis], dramforever[m]1: does regular wlroots/TinyWL/wayland work on your hardware?
<whitequark[cis]> how do i test this?
<whitequark[cis]> dramforever: i think i don't even have the nvidia drivers installed
<whitequark[cis]> wait, do you have the same dell laptop as i do?
<Consolatis[m]1> so indeed looks like a mesa bug, either the caller is wrong or iris_set_damage_region is wrong
<Consolatis[m]1> https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/gallium/frontends/dri/dri_drawable.c#L99 seems to check it though so not sure what is going on. Might already be fixed in master or something
<whitequark[cis]> when was the latest mesa release again...
<whitequark[cis]> > Mesa 25.1.6 Release Notes / 2025-07-16