I have pushed 2 commits that reproduce an incremental compilation bug that leads to a segmentation fault or countless test failures due to out-of-bounds panics:
Steps to reproduce:
First commit, then second:
$ git checkout 958af09d656289d7fe5c01f46c159f7859dd0901
$ cargo test -p ide_completion
<a few tests fail, because the code has a legitimate bug>
$ git checkout 6802a06a824164ff3e274a6bb638c4243cdec654
$ cargo test -p ide_completion
<many more tests fail with "index out of bounds" panics>
$ cargo clean
$ cargo test -p ide_completion
<tests pass>
Second commit, then first:
$ git checkout 6802a06a824164ff3e274a6bb638c4243cdec654
$ cargo test -p ide_completion
<tests pass>
$ git checkout 958af09d656289d7fe5c01f46c159f7859dd0901
$ cargo test -p ide_completion
Compiling hir_ty v0.0.0 (/home/jonas/dev/rust-analyzer/crates/hir_ty)
Compiling hir v0.0.0 (/home/jonas/dev/rust-analyzer/crates/hir)
Compiling ide_db v0.0.0 (/home/jonas/dev/rust-analyzer/crates/ide_db)
Compiling ide_completion v0.0.0 (/home/jonas/dev/rust-analyzer/crates/ide_completion)
Finished test [unoptimized] target(s) in 13.15s
Running unittests (target/debug/deps/ide_completion-68e02c0059fb642f)
running 243 tests
test completions::attribute::tests::no_completion_for_incorrect_derive ... ok
test completions::attribute::tests::test_attribute_completion_inside_nested_attr ... ok
test completions::dot::tests::test_completion_works_in_consts ... FAILED
test completions::dot::tests::test_tuple_field_completion ... FAILED
test completions::attribute::tests::test_attribute_completion ... ok
test completions::attribute::tests::test_inner_attribute_completion ... ok
test completions::dot::tests::test_no_struct_field_completion_for_method_call ... FAILED
test completions::attribute::tests::derive_with_input_completion ... ok
test completions::dot::tests::test_struct_field_and_method_completion ... FAILED
test completions::dot::tests::test_trait_method_completion ... FAILED
test completions::dot::tests::test_trait_method_completion_deduplicated ... FAILED
test completions::dot::tests::test_struct_field_completion_self ... FAILED
test completions::dot::tests::test_no_non_self_method ... FAILED
test completions::dot::tests::completes_trait_method_from_other_module ... FAILED
test completions::dot::tests::test_method_completion_issue_3547 ... FAILED
test completions::dot::tests::test_method_completion_only_fitting_impls ... FAILED
test completions::dot::tests::test_visibility_filtering ... FAILED
error: test failed, to rerun pass '-p ide_completion --lib'
Caused by:
process didn't exit successfully: `/home/jonas/dev/rust-analyzer/target/debug/deps/ide_completion-68e02c0059fb642f` (signal: 11, SIGSEGV: invalid memory reference)
$ cargo clean
$ cargo test -p ide_completion
<a few tests fail, because the code has a legitimate bug>
This reproduces both on stable 1.50.0, as well as rustc 1.52.0-nightly (234781a 2021-03-07).
Apologies in advance for the nightmare that reducing this bug will probably be.
I have pushed 2 commits that reproduce an incremental compilation bug that leads to a segmentation fault or countless test failures due to out-of-bounds panics:
Steps to reproduce:
First commit, then second:
Second commit, then first:
This reproduces both on stable 1.50.0, as well as rustc 1.52.0-nightly (234781a 2021-03-07).
Apologies in advance for the nightmare that reducing this bug will probably be.