00:32 <user21> thank you

11:16 <Ermine> Linux didn't see that partition because rkdeveloptool wrote broken partition entry

11:18 <CounterPillow> oh yeah if you've ever looked at the rkdeveloptool code from rockchip directly (as opposed to any of the forks, e.g. the pine64 one) then you'll learn it's surprising it works at all :D

11:21 <Ermine> Lmao

11:22 <Ermine> We'll look at forks then

12:31 <CounterPillow> I use rockusb from https://github.com/collabora/rockchiprs but the really nice command line tool it comes with as an "example" is undocumented, and it's not entirely feature complete (e.g. partition write/read and eraseflash isn't implemented)

12:32 <CounterPillow> You can install it with `cargo install rockusb --example rockusb --features=libusb` and it'll vomit a binary called `rockusb` into your ~/.cargo/bin

12:41 <qschulz> CounterPillow: also, there is an issue when flashing bmap files with rockusb, somehow the checksum gets broken

12:41 <qschulz> haven't had the time to look at it or produce a simple reproducer :/

12:42 <qschulz> happened on a product we verify the squashfs checksum in U-Boot and it was broken when flashed with rockusb but not with rkdeveloptool

12:44 <diederik> https://xkcd.com/927/

12:45 <mmind00> CounterPillow: should I give you the chance to say something about the slab.h clk thing, or can I just apply things?

12:46 <CounterPillow> mmind00: feel free to apply :) I'm sorry I didn't spot this one earlier

12:47 <mmind00> CounterPillow: no worries :-) ... also my build-test also didn't spot this, so in "our" regular set of build-options/includes/whatever this really does not seem to appear much

12:47 <mmind00> just the test-robot finding that super-special config set to trigger it :-)

12:50 <CounterPillow> qschulz: from sjoerd: if the bmap file has gaps then that's expected.

12:51 <qschulz> CounterPillow: oh that's fair. I assume we do "stupidly" the checksum by simply reading the whole partition

12:52 <qschulz> and I assume squashfs supports holes otherwise my image wouldn't flash faster

12:52 <qschulz> welp, that was an easy find, thanks for the reminder :)

12:53 <qschulz> it probably would work if I did an mmc erase before, but I'm not sure this actually works well (I for sure have issues with some SoCs like RK3399 for example)

12:53 <CounterPillow> yeah, and it sorta defeats the point of bmaps with holes in them ;)

12:54 <qschulz> not really

12:54 <qschulz> mmc erase is much faster than writing zeroes

12:54 <qschulz> blkdiscard is actually almost instant

12:55 <CounterPillow> I see