linux: cp project and device device tree, silence errors

[MilhouseVH]

This gets rid of the following errors when building the kernel:

cp: cannot stat 'projects/Generic/devices//config/*-overlay.dts': No such file or directory
cp: cannot stat 'projects/Generic/devices//config/dt-blob.dts': No such file or directory

Currently the only projects or devices with dts files are Slice and Slice3:

projects/RPi/devices/Slice/config/dt-blob.dts
projects/RPi/devices/Slice/config/ws2812-overlay.dts
projects/RPi/devices/Slice/config/slice-overlay.dts
projects/RPi/devices/Slice3/config/dt-blob.dts
projects/RPi/devices/Slice3/config/ws2812-overlay.dts
projects/RPi/devices/Slice3/config/slice-overlay.dts

With this PR we:

1. Silence the errors when no dts files are present
2. Add support for "common" project-level dts files
3. Avoid accessing the devices directory when $DEVICE is not defined

Perhaps this is overkill for a largely cosmetic issue, but #2 above might prove useful.

[vpeter4]

Is loop really necessary? What about just redirecting error to null?
cp -v ..... 2>/dev/null || :
Instead of 4 lines there are 14 of them :(

[MilhouseVH]

Necessary? No it's absolutely not, but it's the usual method we've used elsewhere so it is at least consistent. :)

As I say it's probably overkill, but it is also perhaps a little bit more "correct" than copying and ignoring/throwing away the error.

Chewitt raised the initial error, so I'll let him decide if he wants this or not, or prefers ignoring the error in which case it can be refactored along those lines while also handling common project files.

[vpeter4]

I just mentioned this because when reading such code it takes more time to think what it does and when something is copied. With one liner cp it is clear immediately. KISS way :)
Not that this has any important factor when running.

[MilhouseVH]

I do agree the one liner is clearer, however I guess it's a choice between brevity (which may cover up potential errors) and being verbose (but safe).

In the past we've tended towards the latter, and I think on the whole it's something I'd favour as I really don't like failing silently unless it's absolutely essential.

[vpeter4]

Well, I think I agree with this. Especially after remembering cp issue with wildcard on 2 folders few days back :-)

Some helper function for such copy would be useful too - loop_cp with two arguments.

[MilhouseVH]

Yes, I think there are a number of places where some helper functions would be useful, this being one such example, particularly as a helper function would automatically handle projects and devices. It's something I intend to tackle. To that end I'd suggest keeping this PR as is, and eventually this type of code will hopefully disappear.

[chewitt]

thanks :)