Where can i report bugs ?

[BloodRagg]

Where can i report bugs ? My kobo has no wifi and i have no github account.

the sftp-server is hardcoded in the dropbear binary as /mnt/onboard/.adds/koreader/sftp-server instead of ./sftp-server.

[NiLuJe]

By design for... (annoying) reasons.

[BloodRagg]

Ahh.. too bad...its the only file left in my .adds folder.

Some other bugs i found

• If /mnt/onboard/.kobo/Kobo/Kobo eReader.conf doesnt exist, koreader crashes.
• If PRODUCT isnt set, koreader crashes.
• If WIFI_MODULE isnt set, koreader crashes (when network is selected in menu).
• If koreader is start without KSM or nickel, screen is upside down for AuraHD
  (fix: case "$PCB" in E606C*) echo 0 > /sys/class/graphics/fb0/rotate;;esac )
• AuraHD doesnt have Ondemand or Interactive scheduler.

Other than that, I'm really enjoying it

[NiLuJe]

Running without Nickel or pickel or ntx_hwconfig is not a supported configuration. There's okreader for that .

(Handling those issues on our own would be fragile, error-prone, and basically reinventing the wheel for, again, an unsupported setup. Incidentally, that's why okreader's hardware support is limited).

As far as CPU schedulers go, it probably doesn't ship with interactive, that's to be expected, but it *should* offer ondemand, though.
In any case, on that front, the startup script is designed to do nothing if nothing sensible is possible on the current HW. Which might indeed be the case on <= Mk.5 devices, given how defensively I coded that after a few rounds of tests on various older models.

[BloodRagg]

Originally Posted by NiLuJe

Running without Nickel or pickel or ntx_hwconfig is not a supported configuration. There's okreader for that .

OKreader ... Ohh nice
Im still using ntx_hwconfig and the koreader.sh suggests otherwise (see reboot clause), but ill take your word for it . BTW this support isn't for me, i just wanted to contribute something back. I will setup a github account and do some PR's if i can find the time.

What is do not understand though is why one would use the variables from rcS instead of relying solely on ntx_config. Because when i look at the default rcS

Originally Posted by NiLuJe

(Handling those issues on our own would be fragile, error-prone, and basically reinventing the wheel for, again, an unsupported setup. Incidentally, that's why okreader's hardware support is limited).

I always reinvent the wheel, then you have something to bring to the table later on

Originally Posted by NiLuJe

As far as CPU schedulers go, it probably doesn't ship with interactive, that's to be expected, but it *should* offer ondemand, though.
In any case, on that front, the startup script is designed to do nothing if nothing sensible is possible on the current HW. Which might indeed be the case on <= Mk.5 devices, given how defensively I coded that after a few rounds of tests on various older models.

Yeah that is weird isnt it, but ill give you the ouput
[root@(none)]# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
conservative powersave userspace performance

[pazos]

Originally Posted by BloodRagg

OKreader ... Ohh nice
Im still using ntx_hwconfig and the koreader.sh suggests otherwise (see reboot clause), but ill take your word for it

Running KOReader by using other launcher is supported (ie: launching KOReader via telnet/ssh) but KOReader expects a working Kobo filesystem (including their "weird" drivers path and their binaries)

Originally Posted by BloodRagg

What is do not understand though is why one would use the variables from rcS instead of relying solely on ntx_config. Because when i look at the default rcS

Repurposing system environment variables is a good thing on a device which should retain compatibility with stock firmware and other 3rd party software.

If you want to make KOReader available on other kind of systems you can create a "koreader-standalone.sh", like we do on BQ Cervantes devices: https://github.com/koreader/koreader...-standalone.sh and get your environment variables there.

[Frenzie]

Originally Posted by BloodRagg

Ahh.. too bad...its the only file left in my .adds folder.

Some other bugs i found

• If /mnt/onboard/.kobo/Kobo/Kobo eReader.conf doesnt exist, koreader crashes.
• If PRODUCT isnt set, koreader crashes.
• If WIFI_MODULE isnt set, koreader crashes (when network is selected in menu).
• If koreader is start without KSM or nickel, screen is upside down for AuraHD
  (fix: case "$PCB" in E606C*) echo 0 > /sys/class/graphics/fb0/rotate;;esac )
• AuraHD doesnt have Ondemand or Interactive scheduler.

Some of those may be easy to fix (especially the first one strikes me as most likely a complete non-issue), but you'd have to do it through a PR and/or giving us a backtrace. (Read: I'm not going to strip my device of Nickel stuff, even if this particular change could probably be tested with a quick rename. )

PRODUCT and to a lesser extent WIFI_MODULE seem to located more in the realm of fundamental necessities. I have considered implementing a "generic fallback Kobo" with a warning message instead of erroring out (so that in some cases a new device could work out of the box) but that'd still depend on the presence some Kobo stuff regardless.

Screen upside down might well be an issue in KSM at the moment with NiLuJe's latest 8-bit change, see here.

[NiLuJe]

(Nope, upside down is because pickel/Nickel hasn't been run and as thus the device is in the boot rota, which we never expect/support).

[Frenzie]

KOReader isn't upside down, but it turns KSM upside down on exit due to fbset.

This line here also needs to restore rotation afterward due to the issue I linked:
https://github.com/koreader/koreader...r.sh#L211-L214

[NiLuJe]

Ha, that's another thing entirely, I missed the gitter link, and was talking about the OP's report only .

Unlike fbset, fbdepth tries *really hard* to keep the rotation as-is, kernel shenanigans included (easier said than done ;p). I'd be curious to have a look at the crash.log on an affected system (since fbdepth will log everything in there)!

EDIT: Ah, and the messed up FBInk prints quite possible make sense in that context, I overlooked that, I'll have to think about how to handle that...

[BloodRagg]

Originally Posted by Frenzie

Some of those may be easy to fix (especially the first one strikes me as most likely a complete non-issue), but you'd have to do it through a PR and/or giving us a backtrace. (Read: I'm not going to strip my device of Nickel stuff, even if this particular change could probably be tested with a quick rename. )

Indeed non-issues, but a robust script goes a long way
I made a PR to pull all ntx_hwconfig values into the environment, its a good base.

Originally Posted by Frenzie

PRODUCT and to a lesser extent WIFI_MODULE seem to located more in the realm of fundamental necessities. I have considered implementing a "generic fallback Kobo" with a warning message instead of erroring out (so that in some cases a new device could work out of the box) but that'd still depend on the presence some Kobo stuff regardless.

Originally Posted by Frenzie

Screen upside down might well be an issue in KSM at the moment with NiLuJe's latest 8-bit change, see here.

Im not running KSM (koreader only)

[Frenzie]

@BloodRagg I phrased that somewhat badly I suppose. I meant complete non-issue in the sense that it'd take no time at all to fix.

[BloodRagg]

@Frenzie np mate

[BloodRagg]

Originally Posted by BloodRagg

• If /mnt/onboard/.kobo/Kobo/Kobo eReader.conf doesnt exist, koreader crashes.

Still crashed on this one ,
Whats the reason behind the custom tar binary ?

[NiLuJe]

busybox's tar doesn't support checkpoints, and/or isn't actually there on some target platforms, and/or doesn't support something else that I can't remember OTOH.