<servili007>
Turl: what is your toolchain of choice?
<Turl>
servili007: for kernel?
<servili007>
si
<Turl>
I have 4.7 linaro unpacked somewhere that I use, it hasn't triggered any bugs as far as I can see
<Turl>
I also use the one on the android tree from time to time
<Turl>
4.4.3
<servili007>
I still have my old build environment saved and could pull it up, but I was trying to make a new one and am seeing bad builds on quite a few different toolchains, possibly other issues as well
<Turl>
hno had issues with 4.8
<Turl>
what gcc version are you using?
<servili007>
4.7
<Turl>
what kind of bad builds btw?
<Turl>
nonbooting, panicing?
egbert has quit [Disconnected by services]
<servili007>
hangs trying to find a screen, a couple of my built modules are coming up bad
egbert has joined #linux-sunxi
<servili007>
my mk802ii was my serial debugger, but it may have met an unfortunate fate. My u1a is just blind testing now. might grab a cubieboard
ganbold__ has quit [Remote host closed the connection]
ganbold_ has joined #linux-sunxi
BJfreeman has quit [Quit: had a good time]
anunnaki has quit [Ping timeout: 245 seconds]
servili007 has quit [Quit: Leaving]
TheSeven has quit [Disconnected by services]
[7] has joined #linux-sunxi
ganbold_ has quit [Remote host closed the connection]
<wowon>
Hi drachensun .. can you give me enlightenment ?
ZaEarl has quit [Read error: Operation timed out]
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
BJfreeman has quit [Quit: had a good time]
rellla has joined #linux-sunxi
n01 has joined #linux-sunxi
tzafrir has quit [Ping timeout: 252 seconds]
paulk-desktop has joined #linux-sunxi
rz2k has joined #linux-sunxi
<rellla>
good morning!
<rellla>
anyone gave xbmc and libhybris a try ?
<rz2k>
rellla: ssvb launched cedarx with hybris
<rz2k>
no idea about the actual frontend
<rellla>
rz2k: he successfully tried vlc
* rz2k
ping techn_
<rz2k>
techn_: why livesuit bsp creates image with any 800mb size of ext4 partition, shouldnt we check the device nand size and create a big one?
<rz2k>
or it will create large image file?
* rz2k
no ideas on how livesuit format handles compression and etc
<rz2k>
s/any/only/
<ssvb>
rellla: unless I got it wrong, hramrach_ tried xbmc with libhybris
hurtigbuffer has joined #linux-sunxi
jelly-home has quit [Ping timeout: 255 seconds]
<mnemoc>
slapin_n1: patrick wood sent an script to the ML after the patch to let the repartitioning tool send the ioctl() to re-read the partition table
hurtigbuffer has quit [Changing host]
hurtigbuffer has joined #linux-sunxi
<hramrach_>
it works with xbmc too
<rellla>
hramrach: yeah. can you document the steps ;) ?
<hramrach_>
except you have to replace the libvecore.so private copy in xbmc installation with a link to libhybris one
<hramrach_>
and it somehow tends to crash
<hramrach_>
never crashed with native but with this hybris it just did
<rellla>
reproduceable?
<hramrach_>
no
<ssvb>
hramrach_: can we list the known problems and possible workarounds at the wiki page?
<hramrach_>
the only known problem is that you need correct android libs
<hramrach_>
I use the one from the commit you gave
<rellla>
ssvb, hramrach: i asked already, if i need the whole /system to make libhybris done? or are the *.so enough? is it needed for compiling? i did not dig into lybhybris, yet...
<hramrach_>
I used whole /system but /syste/lib would probably suffice
<hramrach_>
it's not like you can get part of /system, anyway
hurtigbuffer is now known as jelly-home
<rellla>
maybe i should first try to build libhybris, before asking this questions ;)
<ssvb>
hramrach_: appears that we can, it's less than 2MB of (presumably apache licensed) android binaries :)
<ssvb>
hramrach_: and then we need to add 4 cedarx blobs to the mix
<ssvb>
rellla: libhybris is a really small library and should not give you much problems :)
<hramrach_>
so it is reproducible
<hramrach_>
not sure what triggers the crash
<hramrach_>
either it crashes when playing a video again
<hramrach_>
or when the screen balnks during playback and you try to play another video
<ssvb>
no correlation with some specific file you are trying to play?
<rellla>
ssvb: and if i understood correctly, i have to take care about the binary-version you linked in the wiki. so do i have to upgrade the android/nand - firmware, before i want to mount it to /system?
<mnemoc>
you arlready have commit access to the cedarx-libs repo, right?
<rellla>
hramrach_: seems like a faulty implementation in xbmc if playback works basically.
<rellla>
btw: any idea whats the difference between ioctrl(m_hcedarv, CEDARV_COMMAND_FLUSH, 0); and ioctrl(m_hcedarv, CEDARV_COMMAND_RESET, 0);
<mnemoc>
the first smells purely buffer related while the second about the whole unit
<mnemoc>
(*wild* guess)
<hramrach_>
ssvb: no, seems it can just play one video and then crashes
<hramrach_>
was not problem with native libvecore
<ssvb>
rellla: these are just the cedarx blobs I tried myself :) some other may or may not work
<ssvb>
hramrach_: thanks, that's a good find
* ssvb
wonders if it is possible to reproduce the problem with a small test program without the whole xbmc monster
<hramrach_>
now it crashes on every video, odd
<hramrach_>
depends on the video
<hramrach_>
some break xbmc when they end so it cannot play anthing more som crash it right on start
<ssvb>
hramrach_: but of course this could be also a bug in a cedarx libhybris wrapper
<hramrach_>
worst thing is this is one of the first videos I tested so it must have worked with hybris too
<ssvb>
rellla: a good point, we need to track all these these pesky ABI issues
<ssvb>
rellla: so the CEDARV_IO_COMMAND also keeps changing between cedarx releases :(
ganbold has quit [Ping timeout: 248 seconds]
andoma has quit [Ping timeout: 248 seconds]
<ssvb>
rellla: but wait, doesn't this CEDARV_IO_COMMAND enum belong to some sort of a higher level wrapper on top of libve?
<ssvb>
rellla: we only need libve and could not care less about any other wrapper blobs
<ssvb>
the changes introduced to IFBM_t struct in libve between releases are already a problem, but also tracking ABI for the higher level wrappers is going to be a nightmare
<ssvb>
hramrach_: I always wondered why RAM is clocked at only 360MHz in Mele, maybe memory accesses from VPU are even tougher on hardware than the ones from GPU?
<hramrach_>
ssvb: does not really look like ram corruption
<hramrach_>
too deterministic
<hramrach_>
also the video that causes it is very computation non-intensive
<hramrach_>
yes, looking more at the ram clocks might be good idea but does not really seem like the cause here
<hramrach_>
since changing to a different library solves the issue
<hno>
ssvb, 360 is the default DRAM frequency in the EVB config. What is safe on pretty much any board with any ram.
<hramrach_>
the question is if 480mhz is safe and if it was really extensively tested and if it's even possible to test
<hno>
and yes the cedarx blobls im mele firmware is more tested than what you find in other random android a10 images.
<hno>
it's non-trivial to test what dram frequency is safe. Only known method is trial & error.
<hno>
and since there is not even parity check on memory it's not very easy to reliably detect error.
<rm>
perhaps only the PCB designers know for sure what frequency is safe for a particular board
<hno>
they don't really.
<rm>
remember that e.g. MK802 II uses 408 MHz
<hno>
PCB simulation can give a rough indication however.
<rm>
I take that as a sign of quality vs lousy PCB layout
<rm>
maybe Mele knew it doesn't really work above 360 on their board
<rm>
or just played it safe
<hno>
I think the latter.
<hno>
Compared to most other Alleinner boards I have seen the mele boards generally play in a different leage on stability & safeness, always taking the safe path.
<hno>
quite different from the average tablet manufacturer who generally tries to cut every corner available.
<hno>
but, the original Mele A1000 board is one of the very very first A10 boards and some later designed boards have had access to much more experience in designing A10 DDR layout.
ijc has joined #linux-sunxi
ijc has quit [Client Quit]
ijc has joined #linux-sunxi
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
ibrah has joined #linux-sunxi
Yaku321 has joined #linux-sunxi
<slapin_n1>
Damn inventors if A7HD - why they addign android's /data (nande) to be just ~500MB, and having whole 6G nandi, instead of android's approach with fuse addverticed somewhere?
<slapin_n1>
and the same with sticks
rz2k has quit []
Yaku321 has quit [Disconnected by services]
Yaku-noob has joined #linux-sunxi
ibrah has quit [Quit: Leaving]
vinifm has joined #linux-sunxi
Yaku321 has joined #linux-sunxi
<vinifm>
hi, to read a memory i2c at24cxx must send two bytes of address, but there is no function in kernel space or in i2c_sunxi.c to do this...
<vinifm>
what can I do?
Yaku-noob has quit [Ping timeout: 255 seconds]
ZaEarl has joined #linux-sunxi
ZaEarl has quit [Read error: Operation timed out]
lebernie has joined #linux-sunxi
<hno>
slapin_n1, maybe they made anv verified the layout for 2GB NAND and then upgraded to 8GB without looking into what happened other than that it boots?
<Turl>
slapin_n1: maybe it ran gingerbread once?
vinifm has quit [Quit: Saindo]
<hno>
vinifm, are you sure the kernel i2c framework don't support this?
ZaEarl has joined #linux-sunxi
<Turl>
he quit hno
<Turl>
that sounds like "10 bit addressing", it might not be implemented
<hno>
his problem.
<Turl>
mripard_ may know
<mripard_>
it is implemented in the allwinner driver
<mripard_>
I never tested the 10 bit addressing though
<slapin_n1>
hno, mripard_ he won't need 10-bit addressing for ar24cxx, he don't understand how to do write-read thing and never programmed these chips before with Linux. There is already made driver for at24 in Linux.
<slapin_n1>
just some noncense to that matter
<slapin_n1>
hno: are you really here?
<slapin_n1>
damn, it is so sluggish, on mail a day with 86400 round-trip time is so slow...
<slapin_n1>
hno, Turl: I mean they did not use fuse approach (joint /data + /mnt/sdcard then special fuse module mounted to /mnt/sdcard to emulate fat permissions)
ganbold___ has joined #linux-sunxi
<Turl>
slapin_n1: yeah I know, usually it's because it requires repartitioning if it ran GB previously (where data and sdcard were always separate and UMS was in use)
<Turl>
original A1X devices ran GB
ganbold__ has quit [Ping timeout: 276 seconds]
<hno>
slapin_n1, I am eating dinner.
<mnemoc>
with irc open? shame on you
<hno>
well, I have eaten. Waiting for two of the kids to finish.
<mnemoc>
:)
mlzy has joined #linux-sunxi
mlzy has left #linux-sunxi [#linux-sunxi]
<hramrach_>
ssvb: tried the mele /system and xbmc still crashes on that video
<hramrach_>
not sure how would you reproduce on vlc
<hramrach_>
maybe make a playlist?
<slapin_n1>
hno: can I have your opinion on other mails?
<slapin_n1>
do anybody knows what the fields in DRAM configuration mean?
<slapin_n1>
hno: bon appetit, btw
<hramrach_>
hmm, relllla gone
<hramrach_>
vlc does not crash on the video so it's xbmc specific
<hramrach_>
it just does not do enough set-up/teardown of the cedarx or something
<hramrach_>
but vlc has another problem
<hramrach_>
after a while it gets that striped+noisy output
<hramrach_>
or maybe after it crashes
<hramrach_>
so perhaps does not reset some other bit properly
rellla has joined #linux-sunxi
<hramrach_>
rellla: I tried the crashy video with vlc and it works fine
<hramrach_>
so there is som xbmc specific problem wit setting up libve before playback
<rellla>
hramrach_: so we have to dig into the code. did you try what i meant with CEDARV_COMMAND_RESET / -FLUSH?
<hramrach_>
and it's specific to xbmc+android libve
<hramrach_>
I tried to change the flush to reset but no change
<hramrach_>
it appears that xbmc keeps some context between videos and does not reset something properly for that video to play
ganbold___ has quit [Remote host closed the connection]
<rellla>
i have to see, what vlc resets or does after playing a video and what the difference is to xbmc. It shouldn't be that hard as only android is affected. so the track is there, we have to walk on it...
<rellla>
sadly i have to quit again at this interesting state :( cu later
rellla has quit [Remote host closed the connection]
<slapin_n1>
hno: look at github:slapin/uboot-allwinner branch mtd-nand, I've dissected patches somewhat, now it is easier for review
<slapin_n1>
damnit
<ssvb>
hramrach_: yes, looks like we still have some work to do
<ssvb>
can we somehow get the upstream xbmc developers back on board?