Re: ARM kernel cross compilation issues for iMX53
Wookey <wookey <at> wookware.org> writes:
> +++ Phil Endecott [2012-03-18 20:49 +0000]:
> > The real problem here is a lack of communication. Nowhere is there e.g. a
> > "news" or "blog" type page where someone posts important stuff like "we've
> > decided to change the clocking because $reason in version $version, and you'
> > need to keep u-boot and the kernel in sync because of this; if you don't lik
> > this the changes are $here and $there and you can/cannot revert them
> > with/without breaking other subsystems than the UART". Because this isn't d
> > in public, you, I, and a dozen other people waste hours trying to guess whah
> > happened - time that we could be spending doing something useful, like fixin
> > bugs in Freescale's kernel code.
> > </rant>
> > It would be nice to imagine that some of the other platforms were better. I
> > doubt it, unfortunately.
> Some of the other platforms _are_ better (TI panda/beagle for
> example). Freescale have historically been particularly useless on the
> mainlining-in-a-useful-way, or telling-anyone-anything fronts.
Yes, TI are supposed to be better. The other platforms that I have practical
experience with are VIA (x86) and NVidia (Tegra), who are possibly even worse,
and XScale (in the Intel era) who seemed to be better.
> As Marcin said, in many ways it's easier to let Linaro take the strain
> here as at least Freescale were smart enough to join up so their
> boards are supported. So we can all benefit from the Linaro
> kernel/platform team working out which magic set of stuff actually
> works and save us all doing it individually.
My problem with the Linaro stuff for the iMX53 - also mentioned at
http://chezphil.org/tvcomputer/kernel.html - was that there is no document
saying how much of the Freescale code has actually been incorporated. "Not
enough of it to be useful" was my conclusion, and again I had to waste time
finding that out by trial and error.