belcher_ has joined #maemo-leste
belcher has quit [Ping timeout: 240 seconds]
belcher_ is now known as belcher
Pali has quit [Ping timeout: 252 seconds]
kvw_5 has joined #maemo-leste
kvw_5_ has quit [Ping timeout: 265 seconds]
cockroach has quit [Quit: leaving]
tmlind has quit [Quit: leaving]
tmlind has joined #maemo-leste
cr4y1_ has joined #maemo-leste
<Entitlement> sicelo - [ SUBMITME: iio: accel: st_accel_core: Read mount-matrix from device tree · msm891... ]
<parazyd> We could include that patch if it makes sense
<sicelo> He says, "It's a bit ugly so I've been meaning to rework it and then submit upstream"
<parazyd> ok
<sicelo> I don't think we urgently need it. Actually that's why I didn't open issue in bug tracker, since it's not really a leste problem
<sicelo> For the record, (3), the mount matrix I ended up with in udev to make it work in droid4's natural portrait mode is 0, -1, 0; 1, 0, 0; 0, 0, -1 ... so that's what would need to go into dts once that's working.
<sicelo> Although it's actually probably 0, 1, 0; -1, 0, 0; 0, 0, -1 that should be in dts. Seems the matrix is inverted/transposed in dts, as opposed to udev.
uvos has joined #maemo-leste
Pali has joined #maemo-leste
Jasper[m] has quit [Quit: Bridge terminating on SIGTERM]
cato`[m] has quit [Quit: Bridge terminating on SIGTERM]
kgoetz has quit [Quit: Bridge terminating on SIGTERM]
t_rex has quit [Quit: Bridge terminating on SIGTERM]
mighty17 has quit [Quit: Bridge terminating on SIGTERM]
fLegmatik has quit [Quit: Bridge terminating on SIGTERM]
venji10[m] has quit [Quit: Bridge terminating on SIGTERM]
tvall has quit [Quit: Bridge terminating on SIGTERM]
ollieparanoid[m] has quit [Quit: Bridge terminating on SIGTERM]
afontain_ has quit [Quit: Bridge terminating on SIGTERM]
MartijnBraam has quit [Quit: Bridge terminating on SIGTERM]
venji10[m] has joined #maemo-leste
Jasper[m] has joined #maemo-leste
cato`[m] has joined #maemo-leste
kgoetz has joined #maemo-leste
fLegmatik has joined #maemo-leste
t_rex has joined #maemo-leste
mighty17 has joined #maemo-leste
ollieparanoid[m] has joined #maemo-leste
MartijnBraam has joined #maemo-leste
afontain_ has joined #maemo-leste
tvall has joined #maemo-leste
xmn has quit [Ping timeout: 240 seconds]
inky has quit [Ping timeout: 260 seconds]
inky has joined #maemo-leste
<uvos> sicelo: yup that looks like it
<uvos> btw the physical oritation thing comes from android
<uvos> where untill 4.0 or something you could not apply a matrix
uvos has quit [Quit: Konversation terminated!]
uvos has joined #maemo-leste
<uvos> parazyd: the patch would not do anything for us, as we still need to override it with our matrix that has -y towards the left
<uvos> parazyd: as we expect landscape as the natural orientation.
uvos has quit [Client Quit]
cr4y1_ has quit [Quit: WeeChat 3.1]
mp107 has quit [Quit: The Lounge - https://thelounge.chat]
mp107 has joined #maemo-leste
kek has quit [Read error: Connection reset by peer]
kek has joined #maemo-leste
inky has quit [Ping timeout: 246 seconds]
xmn has joined #maemo-leste
belcher has quit [Quit: Leaving]
belcher has joined #maemo-leste
inky has joined #maemo-leste
jonsger has joined #maemo-leste
jonsger has quit [Ping timeout: 260 seconds]
jonsger has joined #maemo-leste
jonsger has quit [Remote host closed the connection]
jonsger has joined #maemo-leste
pagurus has joined #maemo-leste
ravelo has quit [Quit: Connection closed for inactivity]
uvos has joined #maemo-leste
<uvos> btw directfb has a kms backend
<uvos> and it dosent work either on ddk1.9 :(
<parazyd> heh
<Wizzup> It might make more sense to start X server, perhaps
<uvos> Wizzup: thats what i do rn
<uvos> im compiling a version of driectfb that issues the command mode updates
<uvos> rn
<uvos> but on d4 will take a while
<uvos> anyhow the kms interfaces in ddk1.9 are there only in theory it seams. at least kms blancing works.
<uvos> directfb also hw accel for omap
<uvos> on fbdev
<uvos> probubly not omap4
<Wizzup> does the charging mode need hwaccel?
<uvos> no
<uvos> is just interesting
<uvos> i think it just supports the omap chips with hw blitter
<parazyd> Wizzup: It doesn't make sense to start X because then it's startX->stopX->startX on every boot
<Wizzup> Do we have to start X if we know we do not want to enter charging mode?
<parazyd> It's just a single program now, and all the logic is inside it
<parazyd> So the answer is yes
<uvos> and even if not
<uvos> its pretty silly
<uvos> to start x
<Wizzup> Looks like gentoo removed directfb
<uvos> yeah directfb seams dead
<uvos> but we only have to use it until we can get ddk1.17 to work
<uvos> sdl-on-kms works there
<Wizzup> ddk 1.17, the holy grail :)
<parazyd> Did you document the current status?
<parazyd> And what could be done to help out?
<uvos> no not really
<uvos> yeah ddk1.17 solves a lot of our problems...
<uvos> if only x worked
<uvos> not sure what could be done to help out
<uvos> i just need a very conious block of time
<uvos> and im not sure what avenue to explore
<Wizzup> I suppose being able to trigger the bug with the most simple program
<uvos> so fmg was to the point that on d4 it segfaults because it moves data from one type of texture to another directly
<uvos> that (maybe) is not alowed per spec
<uvos> he wrote an a porgram that repoduces this
<uvos> but i really dident go into this direction at all
<uvos> instead i tryed out pvr on mesa and wanted to try out the hybris glamor on that
<uvos> since it works on d4 with the android pvr blobs
<uvos> so i backported the hybris glamor back to gbm
<uvos> and that fails becaus it cant find a texture format it likes
<uvos> theres where i stopped to do some other stuff for now
<Wizzup> I remember something about that, wrt the texture format, but I suppose that could be many things
<uvos> yeah i dident look at why at all
<Wizzup> iirc fmg ran into that and fixed it somewhere
<uvos> fmg's sticking point is pretty solvable tbh
<Entitlement> Wizzup - [ Commits · maemo-leste-upstream-forks/xorg-server · GitHub ]
<uvos> the porblem is known and you just have to find all the places it happens in glamor
<Wizzup> maybe this is not even for ddk 1.17
<Wizzup> hm
<uvos> but id rather use the mesa patches
<uvos> than straigt pvr
<uvos> Wizzup: yeah also fmg has his progres in a git repo
<uvos> additonally
<uvos> on n900 nothing works for some reason, wrt rendering to gbm textures from gles, fmg mentioned.
<uvos> but idk what the state is there
<uvos> thats pretty mutch all i have
<uvos> not mutch you could say :)
<uvos> if someone wants a crack at it
<uvos> https://github.com/IMbackK/mesa <- this is patched mesa
<Entitlement> uvos - [ GitHub - IMbackK/mesa: Mesa packages with lima and llvmpipe enabled ]
<uvos> https://github.com/IMbackK/pvr-omap4 <- binary patched an packaged (sorta) prv libs
<Entitlement> uvos - [ GitHub - IMbackK/pvr-omap4: mirror of PowerVR OMAP4 libraries ]
<parazyd> Could also be worth trying latest xorg git
<uvos> parazyd: we work ontop of latest git
<parazyd> Along with this, ofc. Not saying latest xorg would solve it.
<uvos> older xorg, forget it
<parazyd> ah you did? ok
<uvos> gles support is lacking even more
<parazyd> Because in our repo it's not latest
<uvos> yeah sec
<Wizzup> uvos: this is not what you ran into right? https://github.com/IMbackK/mesa
<sicelo> Straight pvr works ... but has its own issues - even glmark2 isn't working on Droid 4 with it, unless I've made some mistakes
<Entitlement> Wizzup - [ GitHub - IMbackK/mesa: Mesa packages with lima and llvmpipe enabled ]
<Wizzup> errr
<Entitlement> Wizzup - [ Draft: glamor: Fix rendering of pixmap textures backed by EGLImageKHR image (!55... ]
<uvos> sicelo: glmark works fine
<uvos> sicelo: on ddk1.17 for me
<sicelo> *straight mesa, I mean
<uvos> hmm?
<Wizzup> I am tired today, sorry, I meant this one: https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/568
<Entitlement> Wizzup - [ glamor: Fix rendering of pixmap textures backed by EGLImageKHR image (!568) · Me... ]
<uvos> strait mesa as in llvmpipe?
<uvos> with omapdrm?
<sicelo> Jonathan's mesa
<sicelo> Mesa+pvr
<uvos> yeah
<uvos> it has issues
<Wizzup> uvos: it sounds similar to the texture problem you ran into, I think
<uvos> idk if glmark works
<uvos> Wizzup: no thats not ti
<Wizzup> ok, too bad
<sicelo> You can try. No luck for me so far on postmarketos. Glmark immediately crashes the gpu there ... but it ran with TI's blobs on N900
<uvos> sicelo: ok
<uvos> it runs on full blobs for sure on d4
<uvos> weston and ksmcube at least work fine on the google pvr patches
<uvos> but yeah its buggy
<uvos> gitlab.freedesktop.org/freemangordon/xserver/-/commits/test <- i started here
<sicelo> Yes, also kmscube fine for me on pmos with Jonathan's mesa + pvr. Very reliably too. Only one time I got around 40fps, after trying some UI it didn't like
<uvos> yeah if glmark and other stuff dosent run, maybe we should continue with the blobs
<uvos> idk
<uvos> i gues the legwork with the texture suffeling might be our best bet
<uvos> (as fmg reccomends)
<sicelo> But give it a try too :-p
<sicelo> I'm too new to this stuff and might have done something wrong
<uvos> well i have no idea what im doing either tbh
<uvos> my trade is mechanical engineering
<sicelo> At least you know something about gbm and glamor
<uvos> :P
<uvos> yeah
<uvos> problem is i really dont know anything about gl
<uvos> i only know about kernel interfaces around video/gfx
<uvos> esspecaly with the texture copying problem, it would be great if we had someone with enough knowlage about gles and egl spec to tell us if whats happing in pvr is to spec
<uvos> or if its a bug
<uvos> if its a bug we could complain to ti
<parazyd> I can try getting through to some people if someone can write me what exactly I could ask them
<parazyd> wrt gles/egl
<uvos> is glTexSubImage2D allowed on EGL_KHR_image textures (backed by gbm)
<sicelo> I think I've seen that before ... glTexSubImage2D ... isn't that something missing on pvr? :-/
<uvos> no
<uvos> i think you are thinking about texture clamping modes
<sicelo> Ok. Might be mistaking it with something else
<parazyd> uvos: So depending on a yes or no answer we would know if it's a bug in the blobs?
<uvos> yes
<sicelo> Ah, mistaking it with GL_EXT_unpack_subimage
<uvos> ah yeah
<uvos> i also take the liberty of hosting freemangordons test app here
<uvos> it shows what you have to do to get EGL_KHR_image into glTexSubImage2D without the blobs exploading
<uvos> thanks to this it really istent more than some legwork to do the same in all places in glamor
<uvos> (should at least)
<uvos> with all blobs
<uvos> mesa-pvr has other / maby additional problems
<Wizzup> weren't there also hangs? or no hangs
<uvos> Wizzup: the hang because of the 1bpp textures was solved by fmg
<Wizzup> ok
<uvos> i think it still hangs on exit
<uvos> but i havent tried the latest on the full blob stack
<uvos> i think everyone is now sutably informed
<uvos> maybe we should copy the above log into an issue
<Wizzup> yeah...
<uvos> executing the manual update ioctls after page flip in driectfb dident help :(
<uvos> i hope im doing something wrong...
<parazyd> Also
<parazyd> the IRC channels are on irc.freenode.net (FreeNode). The main channel is now '##OpenGL' (two hashes) and there's also a '##opengl3' channel specific to OpenGL 3.x development.
<parazyd> Perhaps worth dropping by there too?
<uvos> ok i must be doing something wrong
<uvos> DFBARGS=system=fbdev,no-graphics-vt
<uvos> works fine
<uvos> because the vt's cursor causes the display to continue to refesh
<uvos> so it is dfb failing to refesh the display confirmed
Oksana has joined #maemo-leste
<uvos> ok
<uvos> directfb allready supports command mode displays
<uvos> see https://github.com/DirectFB/directfb/blob/e97c8d40ae10585ad10cb55800efcd2ea13fbdf8/gfxdrivers/omap/omap.c#L66 and adding the ioctls to the page filip dosent do anything because they all fail
<Entitlement> uvos - [ directfb/omap.c at e97c8d40ae10585ad10cb55800efcd2ea13fbdf8 · DirectFB/directfb ... ]
<uvos> even OMAPFB_GET_CAPS fails
<uvos> all with Inappropriate ioctl for device
<uvos> tmlind ^^^
<uvos> on the android kernel it wokrks fine
<uvos> seemingly these ioclts are gohne
<uvos> this explains another observation
<uvos> on bionic kexecboot dosent show if you hide the vt cursor
<uvos> this is again because the display wont refresh because the ioctls dont work
<Wizzup> nice find
<Wizzup> are these for omapdrmfb or omapfb?
<uvos> uhh, that might be it
<uvos> omapfb
<uvos> /dev/fb0 is omapdrmfb right?
* enyc meows
<Wizzup> uvos: I think that is just fbdev emulation in omapdrmfb
<uvos> yeah FBIOGET_FSCREENINFO reports omapdrmfb
<uvos> yeah ok
<Wizzup> So that might not support the omapfb specific ioctl's
<uvos> then idk how im supposed to update the display
<uvos> welp
<uvos> i dont think omapdrmfb can work on a command mode display
<uvos> i gues i could open the drm device and call DRM_IOCTL_MODE_DIRTYFB
<uvos> but thats a dirty hack
<uvos> but why dose this test app i wrote work then https://uvos.xyz/maserati/testfb.c
<uvos> makes no sense
* Wizzup needs to sleep - ttyl
<uvos> good night
<uvos> its not waiting for vsync
jonsger has quit [Ping timeout: 260 seconds]
uvos has quit [Ping timeout: 260 seconds]
sunshavi has quit [Read error: Connection reset by peer]
sunshavi has joined #maemo-leste