handbook: add warning about custom kernel configs

[emaste]

Summary

Inform users that custom kernel configs are less well tested, and they should be prepared to participate in the investigation and remediation of any failures.

[emaste]

See also D46870 for a note in GENERIC.

[kp]

documentation/content/en/books/handbook/kernelconfig/_index.adoc

95

introduces or increases?

(Though it's a bit weasel-word-y)

96

Do we need both this and the next line? I'd be tempted to leave this one out. The next line already says "expect to be asked to help debug" and "patches welcome" ought to be implied by our very nature.

[imp]

Maybe overstated the danger a big. The vast majority of customizations are subsetting to omit drivers that never get past probe which is 100% safe.

INET* is really the only common one that causes friction frequently. While other combos can, they are quite rare.

[olce]

documentation/content/en/books/handbook/kernelconfig/_index.adoc

95

A compromise with what @kp suggested?

96

I tend to agree (this line seems redundant with the next one, which has a more neutral tone more appropriate for the handbook).

[emaste]

documentation/content/en/books/handbook/kernelconfig/_index.adoc

95

Yes a bit weasely, but it is also more accurate - there is some small risk of build or runtime issues with GENERIC, and some still small but also larger risk with custom configs.

96

I like this line, because it's very explicit about expectations, and matches what I would hope to see from someone building a custom kernel. I'd rather shorten or remove the next sentence instead.

[emaste]

incorporate feedback from @kp

[emaste]

documentation/content/en/books/handbook/kernelconfig/_index.adoc

96

I removed the second sentence instead (before seeing @olce's comment). If there's a consensus for that one instead of the one I left in I'll make the switch the other way.

[emaste]

Keep only the 2nd of two sentences about participating in the process rather than the 1st.

[olce]

In D46875#1068829, @emaste wrote:

Keep only the 2nd of two sentences about participating in the process rather than the 1st.

I think the generality of the 2nd sentence may fit the handbook better, but no strong opinions. Ideally, it would be completed by a reference to the disclaimer in GENERIC, if D46870 is accepted (if not, then perhaps the 1st sentence would be better here).

[imp]

I like the newer version a bit better. I kinda want some note about yhis being for users willing to test the config.

We should also maybe note over time custom config files will need editing to reflect changes to GENERIC and the project make no specific commitment to compatibility for every customization between versions.