Add example of unintentional None to 2.19 porting guide

[felixfontein]

Ref: ansible/ansible#85605

[felixfontein]

The result can be seen here: https://ansible--2982.org.readthedocs.build/projects/ansible-core/2982/porting_guides/porting_guide_core_2.19.html#example-unintentional-none

[samccann]

@felixfontein @mattclay - looks like the referenced issue suggests some of this is updated in the next core 2.19 version. Is the content in this PR update still valid?

[felixfontein]

I was planning to check whether the PR mentioned there (ansible/ansible#85652) changed something, but only got to do that now. It actually does (the first string_with_none now is a stringified list without the Nones). Since according to ansible/ansible#85605 (comment) the new behavior isn't fixed it's better to wait until that's finalized.

CC @nitzmahone please add a comment here if/when you decide to change something :)

[mattclay]

The fixes in ansible/ansible#85652 and ansible/ansible#85676 resolve the issue with concatenation of None resulting in a literal None in the template output. These fixes are being backported (ansible/ansible#85663) and will be included in 2.19.1.

[felixfontein]

I'll rewrite this accordingly somewhen this weekend.

[felixfontein]

@samccann @mattclay @nitzmahone I've split this up into two sections, since now the behavior is different enough that mixing it into one seems wrong to me. Can you please check the result? Thanks!

[mattclay]

It's worth pointing out that this no longer matters for the common case of passing the template result to a module argument that accepts a string, since None is now automatically converted to an empty string in that case.

See: ansible/ansible#85652

[felixfontein]

Ah, I didn't notice that change. That doesn't work for type=list though, where an empty string as a result of a template in 2.18 and before resulted in [""], and now results in an error. But since [""] was seldom the expected value I guess this doesn't matter much...

[felixfontein]

I added a note: dbe43f4
Does this look good?

[mattclay]

This looks better.

Your point about lists has me wondering if they should default to [] when given None for module arguments, since that's the closest thing to "no list" -- just like we now do for strings defaulting to "" when None is given.

[felixfontein]

Defaulting to [] would be changing behavior (since depending on where None comes from, it used to be [""] before). That sounds like a can of worm that can cause more bad than good. At least for strings, None either came from a template (and used to be "") or was invalid before.

[patchback]

Backport to stable-2.19: 💚 backport PR created

✅ Backport PR branch: patchback/backports/stable-2.19/87d2e26c15ce5f6e329f51ef5c43ddaefaab54b1/pr-2982

Backported as #3013

🤖 @patchback
I'm built with octomachinery and
my source is open — https://github.com/sanitizers/patchback-github-app.

[felixfontein]

@samccann @mattclay thanks for reviewing!