RequireFileExistsRule: enable __DIR__ only for bleeding edge

[staabm]

the rule, which was improved with #5862, did only work before when a opt-in usePathConstantsAsConstantString: true flag was used. I feel this mean only very few users had it in use.

with the mentioned PR, the rule covers more real world use-cases, because the added behaviour does not depend on the opt-in flag.

to play it save, lets enable the newly added logic only in bleeding edge for now

[VincentLanglet]

with the mentioned PR, the rule covers more real world use-cases, because the added behaviour does not depend on the opt-in flag.

Should it behind the same opt-in flag instead then ?

question

[staabm]

Should it behind the same opt-in flag instead then ?

I don't think so, because the opt in flag usePathConstantsAsConstantString is not triggered in bleeding edge.

the new behaviour from #5862 is meant to be generally available in phpstan 3.0 and should be enabled with bleeding edge to verify it works as expected

[VincentLanglet]

Should it behind the same opt-in flag instead then ?

I don't think so, because the opt in flag usePathConstantsAsConstantString is not triggered in bleeding edge.

The new behavior consider __DIR__.'foo' as a constant, which is kinda consistent with usePathConstantsAsConstantString. It could be considered weird to have the check for DIR and not for pathConstants.

the new behaviour from #5862 is meant to be generally available in phpstan 3.0

That's the only concern I have, should it really be generally available ?

Maybe it has to be behing usePathConstantsAsConstantString + bleedingEdge and in 3.0 it will be behind usePathConstantsAsConstantString.

Looking at the doc of https://phpstan.org/config-reference#usepathconstantsasconstantstring ; it explicitely say

When set to true, the magic constants FILE and DIR are resolved to their literal file path values instead of being generalized to string. This enables rules that depend on exact path values, such as checking whether a file referenced by require actually exists.

So I understand by this doc that the #5862 behavior should be behind the flag ?

[staabm]

if I put it behind usePathConstantsAsConstantString + bleedingEdge it will see even less users usePathConstantsAsConstantString alone already have. so it would be even less useful.

from my POV the goal of #5862 is, that RequireFileExistsRule can deliver its real value, even when usePathConstantsAsConstantString is not enabled.

at the time RequireFileExistsRule was implemented, the only way to make it work was with usePathConstantsAsConstantString enabled. IMO its a shortcoming from day 1 of this rule.

the usePathConstantsAsConstantString existed even before

[VincentLanglet]

I don't see why it shouldn't also be behind usePathConstantsAsConstantString but I agree it has to be behing bleeding edge anyway