Thanks for fixing this 🙏 . Could you rebase this?

@nicolas-grekas is it right to have to make ResolveBindingPass aware of the Autowire attribute ? Shouldn't this be handled by the AutowirePass ?

I agree, the bindings pass has no reasons to know about autowiring stuff.
Can't we fix this by changing the order of the two passes?

@nicolas-grekas @stof My initial approach was to fix it in AutowirePass. I don't remember the exact reason, but I couldn't do it like that so I went with this instead.

I can try to change the order of the passes, but I'm worried that might break other stuff. I'll try it sometime today or tomorrow.

It's also not obvious to me what we want here.

What this PR achieves is this precedence, isn't it?

1. autowiring by attribute
2. binding by type+name of argument
3. binding by type only
4. autowiring by type+name of argument
5. autowiring by type only

Dunno if we should account for that but bindings can also be defined by explicit configuration and/or by autoconfiguration.

Dunno how we can make AutowirePass achieve this ordering (or another) given that once bindings are resolved, we don't know if some argument's values came through bindings or via another path (explicit/auto config).

Maybe this approach is the most pragmatic.

WDYT?

Any other opinion here? @stof?

Thank you @HypeMC.