[WIP] kobo-rclone - Kobo wrapper for Rclone

[sherman]

Originally Posted by NiLuJe

@sherman: There's already one (needed for alpha blending to handle transparency in images).

So, err, I may have played a bit with this...

Oops. .

I haven't checked w/ large print enabled, and I unfortunately have the exact same device as you (and @frostschutz), so that doesn't help validating the approach much, but, hey .

Wow!

I was going to try and put something together myself, but hey, I'm not complaining!

I do have access to a Glo in this household. Whether I want to risk messing mum's ereader up though is a slightly different matter...

[NiLuJe]

I updated the attachment after trying it out in large print mode. Tweaked the thresholds to get something working in both modes (so far ).

That doesn't solve the issue of the touch protocol being potentially slightly different on different devices, but, hey, that was fun, at least . And that *can* be overcome with a bit of effort (we do it in KOReader, after all).

EDIT: You can chuck stderr to the void if you just want the final result .

[sherman]

Well, it works (of course) on my H2O.

We will probably need to refine the numbers somewhat to make it more generic hopefully, but the concept certainly seems valid.

For example, for the button height, it might be better to measure it directly, rather than use an estimate percentage. And we probably need a wider range for the button width (15-20 percent?).

Maybe I'll go take a screenshot of mum's Glo.

[sherman]

Alright, so the button width on the Glo is about 143px. That is approx 18.9% of the viewable width (758px). That's on FW 4.7.10413

[NiLuJe]

Oh, yeah, if you can post the screenshot, I can probably cobble something up .

EDIT: Another approach might also be to just scan the next N rows at the first hint of the two buttons, and then stopping if everything's okay, instead of finishing the complete loop and making a wild guesstimate at the end ^^.

[sherman]

Originally Posted by NiLuJe

Oh, yeah, if you can post the screenshot, I can probably cobble something up .

Sure, Here it is.

As an aside, perhaps we should ask a mod to break out this conversation to a new, more relevant thread? Or we could just start a new one?

[sherman]

Originally Posted by NiLuJe

EDIT: Another approach might also be to just scan the next N rows at the first hint of the two buttons, and then stopping if everything's okay, instead of finishing the complete loop and making a wild guesstimate at the end ^^.

That's how I was originally thinking of doing it. No need to continue scanning if you are sure the buttons have been found.

[NiLuJe]

Done .

Breaks early on match, tweaked the (horizontal) match thresholds to be much more lenient (hopefully that should cover everyone), and tweaked a tiny bit the coordinates fudging for the button's center.

'early' is currently set at after .5% lines, which is between 6 and 8 lines, so we still have roughly that again of headroom if it's not strict enough .

It's probably not terribly pretty (4AM brain), but it appears to work ^^.

[sherman]

Originally Posted by NiLuJe

Done .

Breaks early on match, tweaked the (horizontal) match thresholds to be much more lenient (hopefully that should cover everyone), and tweaked a tiny bit the coordinates fudging for the button's center.

'early' is currently set at after .5% lines, which is between 6 and 8 lines, so we still have roughly that again of headroom if it's not strict enough .

It's probably not terribly pretty (4AM brain), but it appears to work ^^.

What on earth are you doing awake at that hour of the morning? *shudders*

I'm trying to figure out the touch protocol. More specifically, where the base coordinate is. My dump of a top-left tap is 48,1038 and top-right is 46,40. That indicates to me that 0,0 is top-right as far as the touchscreen is concerned. Is this correct?

[NiLuJe]

@sherman: Quite possibly, they boot in landscape at whatever 90° CW or CCW that is, and the touch origin may not be rotated when Nickel fixes that on 32bpp FW (because that matches my experience on 16bpp FW where the fb isn't rotated either, fb origin was at the top-right instead of top-left...).

EDIT: I imagine that's what the touch_switch_xy/touch_mirrored_x shenanigans are for in KOReader.

EDIT²: I'll tweak scan_button to output rotated coords, to check if you can use those directly...

[NiLuJe]

There, tweaked it .

After a quick glance at KOReader, it may be a tiny bit more annoying to make right everywhere, but if that works, I'm pretty sure I can cover at least everything except the Touch automatically (that one's a mess anyway, there are seven billion indistinguishable variants that behave differently).

EDIT: H2O²r1 may also be a special snowflake.

[sherman]

@NiLuJe I am in awe at what you've managed to come up with in a few hours.

And here I was earlier just idly musing if it might be an alright idea, and maybe I might give it a go within the next few days...

[NiLuJe]

@sherman: Well, most of it meshes pretty well with stuff I've had to work on recently in FBInk, so, that was easier than it might otherwise have been .

Because that means the groundwork is there, for once, and, more importantly, how it all works is still fresh in my mind .

[NiLuJe]

So far, so good...

Rotated: x=1029, y=582

hexdump -d /dev/input/event1

0000000   51428   23392   49334   00011   00003   00057   00001   00000
0000010   51428   23392   51142   00011   00003   00048   00001   00000
0000020   51428   23392   51174   00011   00003   00050   00001   00000
0000030   51428   23392   51192   00011   00003   00053   01030   00000
0000040   51428   23392   51203   00011   00003   00054   00595   00000
0000050   51428   23392   51215   00011   00000   00002   00000   00000
0000060   51428   23392   51223   00011   00000   00000   00000   00000
0000070   51428   23392   15473   00012   00003   00057   00001   00000
0000080   51428   23392   15757   00012   00003   00048   00000   00000
0000090   51428   23392   15774   00012   00003   00050   00000   00000
00000a0   51428   23392   15785   00012   00003   00053   01030   00000
00000b0   51428   23392   15796   00012   00003   00054   00595   00000
00000c0   51428   23392   15808   00012   00000   00002   00000   00000
00000d0   51428   23392   15818   00012   00000   00000   00000   00000

\o/

EDIT: Tweaked it to hopefully only output the appropriate coords automagically...
EDIT²: And maybe do the right thing on 16bpp FW, although I may have a minor conversion/rounding error somewhere, because I get a slightly different button color...

[sherman]

@NiLuJe I see you've been a busy person while I was asleep...