Unable to test early finalization prevention #45
Comments
jacob-hughes
added a commit
to jacob-hughes/alloy
that referenced
this issue
Jan 19, 2022
The GcSmartPointer trait was a hangover which let the compiler identify `Gc<T>` structs when they were defined as part of a non-rustc library. Now that the `Gc` implementation is part of the standard library, this is unnecessary and adds complexity. This requires the prevent_early_finalization mir-opt tests to be temporarily ignored due to an issue with non-deterministic builds when bringing in the alloc crate across different machines. [1] [1]: softdevteam#45
|
I think we can close this issue? |
jacob-hughes
added a commit
to jacob-hughes/alloy
that referenced
this issue
May 17, 2023
The GcSmartPointer trait was a hangover which let the compiler identify `Gc<T>` structs when they were defined as part of a non-rustc library. Now that the `Gc` implementation is part of the standard library, this is unnecessary and adds complexity. This requires the prevent_early_finalization mir-opt tests to be temporarily ignored due to an issue with non-deterministic builds when bringing in the alloc crate across different machines. [1] [1]: softdevteam#45
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Early finalization prevention is done with the
prevent_early_finalizationMIR transformation. It's not possible to test this with a mir-opt test because diffing MIR generated from the alloc crate leads to different output on different architectures. The crate hash foralloc::Gcis pulled in and that changes depending on which machine compiles the test. Although this isn't necessary to show that early finalization was prevented, right now there's no way to use a wildcard for this particular output when comparing the diffs.The text was updated successfully, but these errors were encountered: