fix: fallback to stat handler when unix.Access fails for virtual file…

[amjadjibon]

#1318

[mvdan]

The fallback logic doesn't seem right. As the TODO says, we need this logic to unconditionally use a handler, rather than use a mix of the real OS and a handler. If we mix both, then the sandboxing is entirely pointless still.

[amjadjibon]

The fallback logic doesn't seem right. As the TODO says, we need this logic to unconditionally use a handler, rather than use a mix of the real OS and a handler. If we mix both, then the sandboxing is entirely pointless still.

Hi, I'm looking at your feedback about the access handler fallback logic mixing real OS calls with handler calls.

I'm wondering if adding a new AccessHandlerFunc to the handler interface (similar to StatHandlerFunc, OpenHandlerFunc, etc.) would be the right approach here? This would let users provide custom access handlers for virtual filesystems, similar to how they can already provide custom stat and open handlers.

Or would you prefer a different direction? The TODO mentions v4, so I want to make sure I'm heading in the right direction before proceeding.

Thanks!