2041 replies to this topic

[Huevos]

 

The data in enigma.info is immutable when loaded into BoxInfo. What Littlesat has broken is something different. He has disabled the ability for BoxInfo to write to SystemInfo. The whole point of BoxInfo is compatibility. If you are going to change the rules now you will not have compatibility with oe-a and third party plugins that are using this feature.

 

What is the point of saying "the source is immutable" (which is itsn't, it is a mutable text file, in that respect the first device driver idea was better), but then make the container reading it, which is used as the source of truth in E2, mutable, and explicitly say it is a feature?

 

There is NO valid reason for any bit of E2 code to need to alter the truth about the hardware, if you allow that, you have just dumped the entire idea about WHY an immutable boxinfo was needed.

 

This sounds very much like the continuous story of OE-A: A good idea, ignoring input from others, then completely make a bodge job of the implementation due to a failure of understanding what it was about.

 

You need to read the code. The values coming from the text file are read into BoxInfo as immutable.

 

What has been broken is the ability to modify SystemInfo.

 

For BoxInfo discussion I have started a new thread here.  https://forums.openp...nfo-discussion/

 

Maybe someone can move the posts from here.

Edited by Huevos, 30 December 2023 - 14:08.

[DimitarCC]

Well the problem of the infobar is that it uses 2 big png images with alphablending.

Maybe use images only in corners will fix the issue (like it is in second infobar)

I think so, yes.

Ok i will rework the infobar in this way and see how it will goes.

[littlesat]

Maybe consider to build the stuf in sections. So four corners that are small and need the alpha blending and then draw the rest that does not need alpha blending:
Thanks for the efforts…

Edited by littlesat, 30 December 2023 - 14:47.

[WanWizard]

Maybe consider to build the stuf in sections. So four corners that are small and need the alpha blending and then draw the rest that does not need alpha blending:

 

That is exactly how the template Screen_Base is defined.

 

btw, I've just flashed a Zgemma H9, and it doesn't have the problem, so it is not that HiSilicon SoC's don't support alpha blending and/or hardware accelleration, but that it is not (properly) implemented in the drivers.

From that I can see, both the H9 and the Dual 4K are built with

EXTRA_OECONF_append_pn-enigma2 += " --with-alphablendingacceleration=always --with-blitaccelerationthreshold=250 --with-fillaccelerationthreshold=190000"

but it looks like this is more copy/paste instead of it actually be supported.

 

I'll see what happens if I compile E2 for the Dual 4K without this.
 

[DimitarCC]

Maybe consider to build the stuf in sections. So four corners that are small and need the alpha blending and then draw the rest that does not need alpha blending:
Thanks for the efforts…

 

 

Yes. that is the way it works for other screens. only corners are alphablended images and the rest is plain eLabel with background color.

 

 

 

btw, I've just flashed a Zgemma H9, and it doesn't have the problem, so it is not that HiSilicon SoC's don't support alpha blending and/or hardware accelleration, but that it is not (properly) implemented in the drivers.

From that I can see, both the H9 and the Dual 4K are built with

EXTRA_OECONF_append_pn-enigma2 += " --with-alphablendingacceleration=always --with-blitaccelerationthreshold=250 --with-fillaccelerationthreshold=190000"

but it looks like this is more copy/paste instead of it actually be supported.

 

I'll see what happens if I compile E2 for the Dual 4K without this.
 

 

 

That is interesting.... If H9 doesnt have the issue but Dual 4K have then should be the drivers....

Yes. that is the way it works for other screens. only corners are alphablended images and the rest is plain eLabel with background color.

[littlesat]

It is a pitty to change it because drivers have issues…

[WanWizard]

That is interesting.... If H9 doesnt have the issue but Dual 4K have then should be the drivers....

 

most likely. Now flashing the Dual without acceleration enabled, see if it makes any difference.

[WanWizard]

It is a pitty to change it because drivers have issues…

 

Always the same problem with OEM resellers, they don't have the developers, the skills, or the willingness to spend the money to pay them.

 

Sofar it looks like it is limited to the list @Ev0 mentioned: Qviart Dual, the SF8008 / Trio4k / uSytm4kpro and the Multibox.

[WanWizard]

most likely. Now flashing the Dual without acceleration enabled, see if it makes any difference.

 

It doesn't. If anything it even looks marginally slower.

[DimitarCC]

I see... Well then they have not correct or not developed enough drivers as you mentioned.

[Ev0]

 

most likely. Now flashing the Dual without acceleration enabled, see if it makes any difference.

 

It doesn't. If anything it even looks marginally slower.

 

No idea if this info helps or not.

But I just put the OE-A drivers into OpenPli 9.0 and it didn't change anything.

 

[WanWizard]

I didn't think so, I'm pretty sure it is something worked around in the E2 cpp code, and not fixed in drivers...

[DimitarCC]

Happy New Year to everyone!

 

So i have already changed the skin with the alternate rendering of infobar. Currently in the repo so if someone like to test on HiSilicon box just have to get the diffs...

 

There are only 2 small background images with alphablending now in infobar. So i hope it will be fast.

Edited by DimitarCC, 2 January 2024 - 09:40.

[littlesat]

Thanks!

Edited by littlesat, 2 January 2024 - 09:38.

[WanWizard]

So i have already changed the skin with the alternate rendering of infobar. Currently in the repo so if someone like to test on HiSilicon box just have to get the diffs...

 

There are only 2 small background images with alphablending now in infobar. So i hope it will be fast.

 

Thanks. Just build a new image, it is a MASSIVE difference !

[DimitarCC]

Cool! I hope now is almost instant show up of the infobar...

[WanWizard]

Yes, it is, I no longer see any build-up.

[littlesat]

A pitty that it is needed... but the result is really wonderful!!!! 

[DimitarCC]

So i have created a pull with recipe for boxlogos from my repo

 

https://github.com/O...-core/pull/1225

 

However i am not sure is it on correct place so check it out.

 

Later a dependency in the skin can be added.

[WanWizard]

I'll close that PR if you don't mind.

 

I was already busy with it, but it needs quite a bit of customization due to the fact we build images for multiple boxes, so we need image selection at runtime, not at build time.