Isn't this exactly what I've tried here? #8512 It didn't work because it indeed broke stuff downstream.

Instead of filtering, I thought we agreed to turn them into fallback instead.

Ah yes I was not sure if you meant to make these rules fallback too. I just updated the PR but there's a test failing:

;; More complex test

(rule
 (target into+ignoring)
 (mode (promote (into subdir)))
 (action (with-stdout-to %{target} (echo "hello"))))

(rule
 (target into+ignoring)
 (enabled_if %{ignoring_promoted_rules})
 (action (copy subdir/into+ignoring into+ignoring)))

I think it's invalid now, but it's probably a smaller repro of some existing package. WDYT?

I think it's invalid now, but it's probably a smaller repro of some existing package. WDYT?

Isn't this basically a manually created fallback rule?

In any case, given that this variable exists (though it seems to be used) I feel like this change is big enough that perhaps we should save this all until 4.0?

Agree. Let's keep #8706 though for 3.11.

What's the rush to get it out for 3.11? The bug fix is basically incomplete so this fix helps nobody still. Moreover, it introduces a confusing behavior

What about generated opam files that now fail because of #8518? Do we call that an intended breakage?

I'm reverting that as well.