Stream: compiler development

Topic: memory safety for executing pure package code

view this post on Zulip Anton (Sep 04 2026 at 18:03):

I think @Richard Feldman once talked about this LLVM stance "vulnerabilities in clang, clangd, or the LLVM optimizer in a JIT that are triggered by untrusted inputs are not treated as security-sensitive." as an argument for not using ReleaseSafe in the Roc compiler.

I was thinking that our separation with pure packages changes this trade-off. What kind of memory safety should we strive for in order to prevent malicious pure Roc packages from emitting specific effectful code by exploiting vulnerabilities in the compiler?

Last updated: Sep 24 2026 at 15:59 UTC