[rectang]

At first glance, watt looks like a substantial improvement that would close the door on arbitrary code execution by proc macro crates. Yes, please! While this may not solve the general problem of package identity validation, it closes a Rust-specific hole that hopefully doesn't need to exist.

Now if only `build.rs` could be nerfed...

[duped]

build.rs is particularly useful for Rust because it is routinely used to compile C/C++ object files as a previous step, which is crucial to having solid Rust to C/C++ FFI.

It is no different from a ./configure script, or other prebuilt script. Lots of builds require these, and "nerfing" it just makes building Rust harder. Cargo is already a crippled build system that requires extensions like cargo-make to be useful. Getting rid of something so fundamentally required by modern software with no standard fallback would be a massive blow to the ecosystem.

I really am not convinced that there is anything "scary" about a build.rs file - other than that standard tools like rust-analyzer find it sane to run external code during initialization. Your language server shouldn't be coupled to the build system and require it to run!

(And yes, Cargo is a build system - it's just a bad one)

[rectang]

sigh, probably "nerfed" wasn't the greatest choice of words... I'm writing such an FFI crate right now, and I use a `build.rs`. I can still wish that the package management system didn't have to fall back to running arbitrary code, or that there was some way to sandbox that code. That would make it easier for people to trust my crate!