[Cervantes]: waveforms & flags

[pazos]

Removed a bunch of unused structs & kernel code. Not really needed but makes the file easier to read/understand.

Use same MXCFB_SEND_UPDATE because MXCFB_SEND_UPDATE_OLD was borked and made all legacy devices stupid enough to not update the screen 😭

Add a bunch of epdc flags from Cervantes 2013 - Cervantes 4 kernels.
Rename is_new to is_reagl to be a little more descriptive.

This fixes #754 (in particular: #754 (comment))

[NiLuJe]

At first glance, LGTM ;).

A few stray FYI comments:

• I usually keep the mxcfb includes as close (and as messy) as upstream as possibly simply to make diff'ing against newer versions easier. Granted, this is mostly because lab126 kernels are a moving target, this may not be as interesting for "quieter" brands ;).

• I've never seen anything requesting the "D" variants of the REAGL waveform directly. That's not to say no-one did or does, just my personal experience ;).

[NiLuJe]

And I had a vague inkling I was talking out of my ass on that last point, and yeah, I was: the Aura was the special snowflake doing just that, don't mind me :D.

[pazos]

I've never seen anything requesting the "D" variants of the REAGL waveform directly. That's not to say no-one did or does, just my personal experience ;).

Technically I had no idea, stock reader relies on auto selection most of the time (on the ui) but is a bit messy because it forces full update mode almost everywhere. While turning pages auto is used too and I 'm clueless about which wfm did the kernel autoselect.

GLD16 waveform seems nice with EPDC_FLAG_USE_DITHERING_Y1 on my Cervantes4 for non flashing ui refreshes, but seems non standard so I leave that for myself and use auto here, as most of other devices do. (Using AAD here just slows the UI)

GLR16 works nice too for partial updates (it is a bit faster than GLD16 w/ the same level of ghosting artifacts) but provides somewhat a clunky experience turning pages.

[NiLuJe]

I don't remember the details of how the Aura handled things, but on Kindle, a REAGL update must be _FULL, or weird things happen ;).

Note that even like that, they do NOT flash, so a flashing update switches to GC16 ;).

---

But that's on Kindle, as I said, don't remember how the Aura does it. But I think I got someone to send me strace logs, so our current behavior should somewhat match Nickel in that respect, if you want to cross-ref ;).

[pazos]

I don't remember the details of how the Aura handled things, but on Kindle, a REAGL update must be _FULL, or weird things happen ;).

@NiLuJe Now I understand your verbosity levels 😄
https://github.com/koreader/koreader-base/blob/master/ffi/framebuffer_mxcfb.lua#L565-L567

So, yeah, setting waveform_reagl will do the full update trick, like in the Kobo Aura. The trick seems not required in Cervantes but is used to have a working night mode (it is ghosting like hell otherwise)

[NiLuJe]

Yeah, I'm pretty much doomed to forget 95% of it otherwise ;p.

[Frenzie]

I usually keep the mxcfb includes as close (and as messy) as upstream as possibly simply to make diff'ing against newer versions easier. Granted, this is mostly because lab126 kernels are a moving target, this may not be as interesting for "quieter" brands ;).

In a similar vein, I think it's easier to quickly experiment a little if you don't have to mess around as much with header definitions.

[Frenzie]

lgtm

[pazos]

@Frenzie: Can this PR be merged ?. I'm not in a hurry but I'd like to start using OTA updates without building them myself 😃

[Frenzie]

I wasn't sure if @NiLuJe had any more comments. :-)

[NiLuJe]

Nope, LGTM ;).