On Sat, Dec 26, 2015 at 5:38 PM, Paul Boddie paul@boddie.org.uk wrote:
On 2015-12-26 17:28, Luke Kenneth Casson Leighton wrote:
thanks paul. i've reached out to the linux-mips community, i contacted them last year and at least 2 people expressed an interest in helping with board bring-up and kernel development. jz4780 is very similar to jz4775 and there have been several commits which cater for both processors.
Last time I looked, it seemed that the jz4780 had benefited from the work done to support the MIPS Creator CI20, which had probably benefited from the ongoing work done with the jz4770 (in the GCW-Zero and various handheld consoles, as I understand it), so the jz4775 shouldn't bring too many surprises, especially since it comes without the PowerVR nonsense. I guess that even the device tree support in the jz family is fairly complete now, too.
good timing. it's still an "unknown" however, so my first instinct is to use "known-good" as a means to confirm the working hardware. that means using the code released by ingenic, which is based (iirc) on linux kernel 3.0.8. *AFTER* the hardware has been confirmed operational, *THEN* the hardware may in turn be used to confirm that the [latest] linux kernel tree works correctly.
a very simple very hard lesson learned there: go ONLY from ONE known-good to another. UNDER NO CIRCUMSTANCES attempt to introduce two (or more) highly-complex inter-dependent unknowns.
this was why the LCD took almost two maybe as long as three months to get operational, because there were no less than *five* separate and distinct unknowns in the chain, including the linux kernel tree (sunxi-3.4-cb), script.fex file, LVDS cable (which had to be made up by hand), LCD itself, backlight circuit, backlight and lcd GPIO, and the TI LVDS converter IC. and the CPU Card. how many's that so far? 8 or 9 separate unknowns, each inter-dependent?
l.