Are apcu, redis, and memcached needed if caching.apcu, caching.redis, and caching.memcached are set to false?

Hmm. Good question. Maybe not. With the new change, we could definitely make these extensions dependent on the options, and it's a good idea if it works.

It seems that the memcached option is actually never used for anything. Is that right?

My impression is that in this commit, the extension setting was changed from optionally adding the three caching extensions to always adding them: ed20aae#diff-80ddb2d38f0f6e01125ccb8683381d7be20be1df3dce01e5c200dd74f5173623

So probably the right way forward now is to make them optional again, since this is easy now.

No. See nixos/tests/nextcloud/with-mysql-and-memcached.nix.

Are apcu, redis, and memcached needed if caching.apcu, caching.redis, and caching.memcached are set to false?

I don't know. And I'd be especially cautious with apcu as this is something that's something where people often assume that it's "just available".

Is it really worth making this module even more complex to save the closure footprint of a single php extension though?

In that case I would recommend removing those options if they do nothing.

No. See nixos/tests/nextcloud/with-mysql-and-memcached.nix.

What do you mean? That test manually sets up memcached.

Is it really worth making this module even more complex to save the closure footprint of a single php extension though?

I don't know. Does the PHP extension for redis pull in redis-the-service/package as an additional dependency? Then it would be worth making it configurable. If it's just a few small PHP files, then I agree with @aanderse that the options should be removed, and the documentation adjusted that these extensions can always be assumed to be available.

@Ma27 how shall we go forward?

Have you confirmed whether this is actually fine?

I took this commit from @aanderse . I don't know whether it is correct. It worked on my server, but I didn't tweak this setting.

@Ma27 as you can see it is applied twice. I looked and confirmed duplicate config was generated.

This doesn't even evaluate since there's a missing ++.

It may be better to put ++ before a line, then it's also easier to comment a single line out during debugging.

Better?

<xref linkend="opt-..." />.

That's confusing. The rest of the document uses the format I took. I don't know what the difference is.

The text of the link is generated automatically by <xref />, while it needs to be specified with <link>…</link>. This is used throughout the manuals.

Fixed