hi everyone, i’m Abhinav ( paradoxicalguy (abhinav ) · GitHub ), a student from mumbai.
i contribute to the Rust compiler, mostly on reproducible builds: making sure the same source produces the same binaries no matter where or how it’s built. currently 23 merged prs across rust-lang repo
FOSS United awarded me a 6-month fellowship under Season of Commits for this work. i’ll log progress in this thread roughly bi-weekly or monthly.
recently i ran an A/B reproducibility test on aarch64 Linux and found 12 of 74 artifacts differing. the differences disappear after normalizing checkout paths and excluding embedded LLVM bitcode, so the remaining non-determinism is mostly path leakage.
approach: triaging existing reproducibility issues, starting with older ones, and add regression tests w each fix.
milestones
by month 3:
- deterministic rlib object names [ make rlib object names deterministic by paradoxicalguy · Pull Request #165202 · rust-lang/rust · GitHub ]
- reject Reborrow impls on types with &'static mut fields [ reject Reborrow impls on types with &'static mut fields by paradoxicalguy · Pull Request #165006 · rust-lang/rust · GitHub ]
- land the follow-up MSVC regression test PR: [ tests: cover proc-macro metadata path remapping by paradoxicalguy · Pull Request #166156 · rust-lang/rust · GitHub ]
- fix: non-deterministic temporary object paths in linker maps [ fix: non-deterministic temporary object paths in linker maps by paradoxicalguy · Pull Request #165971 · rust-lang/rust · GitHub ]
by month 6:
5. further reproducibility fixes, including with its regression test
6. a public write-up of the findings