Re: [PATCH 4/4] dmaengine: fsl-edma: integrate TCD64 support for i.MX95

[Frank Li]

On Fri, Nov 10, 2023 at 04:10:46PM +0100, Krzysztof Kozlowski wrote:
> On 10/11/2023 15:59, Frank Li wrote:
> >>>
> >>> Signed-off-by: Frank Li <Frank.Li@xxxxxxx>
> >>> ---
> >>
> >> Three kbuild reports with build failures.
> >>
> >> I have impression this was never build-tested and reviewed internally
> >> before posting. We had such talk ~month ago and I insisted on some
> >> internal review prior submitting to mailing list. I did not insist on
> >> internal building of patches, because it felt obvious, so please kindly
> >> thoroughly build, review and test your patches internally, before using
> >> the community for this. I am pretty sure NXP can build the code they send.
> > 
> > This build error happen at on special uncommon platform m6800. 
> 
> Indeed csky and alpha are special. Let's see if LKP will find other
> platforms as well.
> 
> > Patch is tested in imx95 arm64 platform.
> 
> That's not enough. It's trivial to build test on riscv, ppc, x86_64 and
> i386. Building on only one platform is not that much.
> 
> > 
> > I have not machine to cover all platform.
> 
> I was able to do it as a hobbyist, on my poor laptop. What is exactly
> the problem that as hobbyist I can, but NXP cannot?

There are also difference configs. I think 'kernel test robot' is very good
tools. If there are guide to mirror it, we can try. It is not neccesary to
duplicate to develop a build test infrastrue.

The issue is not that run build test. The key problem is how to know a
protential problem will be exist, and limited a build/config scrope.

Even I have risc\ppc\x86_64 built before I submmit patch, still can't
capture build error if I missed just one platform mc6800.

For `readq` error also depend on the configs.

Actually, we major focus on test edmav1, .... v5 at difference platforms
before submit patches. 

> 
> Best regards,
> Krzysztof
>