Changelog entry for the change of dune cache default (#10710)

[ElectreAAS]

No description provided.

[ejgallego]

A couple of comments:

• should doc be updated in this PR, or in a separate one?
• IMHO the most important bit is to explain how uses can re-enable old behavior, I am aware of users in the wild relying heavily on cache + safe user-side rules. I guess they all need to do is to update the ./config/dune as to enable the cache for the user rules, right?

[ejgallego]

Thanks for the update @ElectreAAS !

I have a last question, is it possible with the new code to recover the old behavior when the user had used (cache enabled) in the config file, without using any env variables?

[Khady]

to be clear, this is about caching the output of such rules right? Not the rules themselves

[ElectreAAS]

I don't really know what caching a rule without caching its result would even mean. So yes I believe this is about caching outputs.

[Leonidas-from-XIV]

I have a last question, is it possible with the new code to recover the old behavior when the user had used (cache enabled) in the config file, without using any env variables?

No, not with the current code, you need to explicitly opt-in to caching the user rules.

[rgrinberg]

is it possible with the new code to recover the old behavior when the user had used (cache enabled) in the config file, without using any env variables?

As Marek mentioned, it's not possible at the moment. If you'd like to add a config option for this in addition to the env var, you are welcome to.

[ejgallego]

As Marek mentioned, it's not possible at the moment. If you'd like to add a config option for this in addition to the env var, you are welcome to.

Thanks folks for the confirmation.

I think indeed an option to enable this in ./config/dune should be added, otherwise quite a few users I know (virtually all Coq devs + many users) will regress as they have a setup where they enable the cache in the config file (their rules are safe), however it is far from trivial for them to switch to using environment variables.

I guess that would be a regression, but likely this setup did fall under the "experimental" category, right?

[Leonidas-from-XIV]

I am aware of the discussion happening in #10710 but I think this PR accurately describes the current state of main. If any changes will be done that will happen in a separate PR so I think merging this one in the meantime is sensible.

[nojb]

I am aware of the discussion happening in #10710 but I think this PR accurately describes the current state of main. If any changes will be done that will happen in a separate PR so I think merging this one in the meantime is sensible.

The discussion in #10710 is converging, so you may want to hold off before merging this one. See #10944.

[ElectreAAS]

#10944 has been merged, and it includes a correct changelog.
Closing this one.