[Gta04-owner] We need an Android kernel

[Dr. H. Nikolaus Schaller]

Am 29.07.2013 um 14:56 schrieb Paul Kocialkowski:

> Le lundi 29 juillet 2013 à 14:30 +0200, Dr. H. Nikolaus Schaller a
> écrit :
>>> I will get back to working on Android devices given that the situation
>>> hasn't improved.
>> 
>> Well it is work and isn't improving by observing...
> 
> Yes I understand that, though I'm not very efficient when it comes to
> working with the kernel, so watching and testing is nearly as much as I
> can do.
> 
>>> I plan to get back to work on Android userspace for
>>> gta04 as soon as there is a working Android kernel.
>>> 
>>> Either trying to make an Android kernel work on gta04
>> 
>> After the recent discussions I think it is the more troublesome approach
>> to make some (old) Android kernel (with its own bugs and deficiencies)
>> work on the GTA04.
> 
> I wrote the words in that order, but I don't really mean we should
> prefer one approach (making gta04 work on Android kernel) over the other
> one (making Android work on gta04 kernel).
> 
> Moreover, I think you convinced me that using recent kernels with
> Android things enabled is the way to go.
> 
>>> or waiting for one
>>> is a considerable waste of my time, so I prefer to leave that to
>>> competent people and focus on what I do best, which is working on other
>>> Android phones.
>>> 
>>> It seems to me like what happens next is up to you: either some real
>>> kernel works gets and we can move on, or Android on GTA04 will stay in
>>> this unusable status and people will lose interest as the hardware will
>>> get older and older.
>>> 
>>> I think the biggest issue we have here is the DM37x platform. Things
>>> would have been considerably better with an OMAP 3630, and the best
>>> choice of all would have been using one of Allwinner's platform, such as
>>> the A10 or the A13, which have a very active community of
>>> U-Boot/Linux/Userspace developers.
>> 
>> Did you read my mail this morning? It looks very good with the 3.11-rc2 kernel
>> and we just need to know what is not working, i.e. real bug reports [1].
> 
> Actually, I didn't read it before now. I thought, given the title of the
> thread, that it was still about how to get 1Ghz working.
> So this is great new! If suspend/resume works ok, that's the way to go.
> I'll ask Denis to test it soon then.

Yes, would be interesting to know! Since suspend/resume had worked on
the 3.7 kernel (at least for non-replicant OS like QtMoko) and the 3.11
kernel was not a complete rework, I am quite sure that it works. If not
we have to define test cases and analyse.

> 
>> And we are really willing to solve these issues - if someone tells us what
>> is not working to run Replicant on top. If there are no specific reports we
>> can't fix anything...
>> 
>> BTW: the OMAP3630 IS the DM3730. There was never a public chip
>> called OMAP3630 - this was the working title what was sold as DM3730.
>> 
>> Like the OMAP3430 sold to Nokia and Palm was made an OMAP3530 in
>> the public.
> 
> My point is that there are no Android phones sold with DM37x, while
> there are a few other OMAP3 out there, which have a proper 3.0 kernel
> from TI or from the manufacturers. With that granted, adding gta04
> support would have been dead easy.
> 
> The BeagleBoard-xM Android kernels all come with missing suspend/resume
> support as far as I could see, or then they use the very old 2.6.37
> kernel.
> 
>> And if you think there is lack of community support - it isn't. There is the
>> BeagleBoard XM using *exactly the same chip* as the GTA04. So each bit
>> fixed in the mainline kernel tree for the BeagleBoard XM helps the GTA04
>> and vice versa. 
> 
> It seems very much like it, given that suspend/resume for DM37x seems to
> get better and better at every mainline release.
> 
> So maybe there is hope :)

Yes, I am sure. And we will even start to push more things to mainline...