Fix rank mismatch panic during type generalization with var reassignment #8850
Conversation
When a type variable in var_pool at rank N was unified with a variable from a nested scope at rank N+1 (after being added to the pool), and that nested scope was subsequently popped, the variable would resolve to a descriptor still at rank N+1. This caused a panic when trying to add it to tmp_var_pool during generalization. The fix caps the effective rank at the rank being generalized and updates the descriptor's rank to match, treating the higher rank as stale since its scope has already been popped. Co-Authored-By: Claude <claude-opus-4-5-20251101@anthropic.com>
- Move test/fx/issue8848.roc to test/cli/issue8848.roc to avoid platform overhead - Add test to roc_subcommands.zig instead of fx_platform_test.zig - The CLI test runs `roc check` without needing to build a platform Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The test now correctly detects when the compiler panics by checking for signal termination (SIGABRT) instead of just checking the exit status. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
|
if you reassign a mutable var in a higher rank (eg inside a function), it should unify with the root mutable variable at the lower rank. this unification should lower the reassignment variable, making it not eligible for generalization when that higher rank is popped. at that point, it should be moved into the lower rank. that said, it sounds like that is not happening, hence this bug! i'm gonna pull this down and investigate, as i'm pretty sure the |
|
hmm actually, i think this may have to do with value restriction! when we exit a nested scope and it's not a lambda, we simply pop the var pool rank, but actually we need to move those variables to the rank rank above or, we can not push a new rank at all if the nested expr isn't a lambda. will look into |
The test only needs to verify that type checking with mutable var reassignment in nested scopes doesn't cause a panic. A snapshot test is the minimally-expensive test type that still exercises the fix. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Fixes #8848
This PR fixes a panic that occurred when mutable variables (
var $index) were reassigned within nested scopes like match branches. The panic message was "trying to add var at rank 2, but current rank is 1" during the generalization phase of type checking.The root cause: When a type variable in var_pool at rank N was unified with a variable from a nested scope (rank N+1), and that nested scope was subsequently popped, the variable would resolve to a descriptor still at rank N+1. This caused a panic when trying to add it to tmp_var_pool during generalization.
The fix:
Co-authored by Claude Opus 4.5 (claude-opus-4-5-20251101)