Rust: Upgrade to rust-analyzer 0.328#21714
Conversation
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…scripts - Fix Meta usage: Meta is now abstract with subtypes (PathMeta, KeyValueMeta, TokenTreeMeta, etc.) - Fix FormatArgsArg: getName() replaced by getArgName() returning FormatArgsArgName - Add upgrade script (old → new) and downgrade script (new → old) - Update Definitions.qll, PathResolution.qll, BadCtorInitialization.ql, FormatTemplateVariableAccessConstructor.qll Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
tausbn
changed the title
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
Re-run cargo fmt after clippy --fix to ensure consistent formatting. Re-run codegen to update generated file tracking for MetaImpl.qll. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Setting proc_macro_processes to 0 causes an index-out-of-bounds panic in ProcMacroServerPool::new when the proc macro server is enabled. Use the same defaults as rust-analyzer itself (1 each). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Remove format_args_arg_names from upgrade delete list (table exists in both schemas with different columns) - Accept updated .expected files for schema changes Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Exclude macro-expanded and attribute paths from PathResolutionConsistency (tokio::main and similar attribute macros resolve to multiple proc macro fns) - Exclude "macro expansion failed" warnings from ExtractionConsistency (compile_error! and undefined macros are expected to fail expansion) - Update pre-existing consistency expectations (net multipleResolvedTargets) - Update type-inference.expected for new RA results Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
tausbn
force-pushed
the
tausbn/rust-upgrade-to-rust-analyzer-0.0.328
branch
from
c990d17 to
ac27c20
Compare
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
geoffw0
left a comment
geoffw0
left a comment
There was a problem hiding this comment.
Copilot wasn't able to review this pull request because it exceeds the maximum number of files (300).
Yeah, I feel that, though I suspect some of these files visible in the diff are generated code not marked as such. I've skimmed through all the changes anyway, but I'll have to rely mostly on testing (automated and manual) to build confidence here.
Can I get confirmation:
- that the Copilot-written changes have been reviewed by the author?
- that the upgrade / downgrade scripts have been tested?
| trait_alias_names.rel: delete | ||
| trait_alias_type_bound_lists.rel: delete | ||
| trait_alias_visibilities.rel: delete | ||
| trait_alias_where_clauses.rel: delete |
There was a problem hiding this comment.
Why do both the upgrade and downgrade scripts just delete a bunch of relations? Don't we have to do anything for the new relations?
There was a problem hiding this comment.
this indeed seems like a placeholder upgrade script, we probably need something better.
| not exists(ew.getLocation()) or ew.getLocation().fromSource() | ||
| (not exists(ew.getLocation()) or ew.getLocation().fromSource()) and | ||
| // macro expansion failures are expected for macros like compile_error! and panic! | ||
| not ew.getMessage().matches("macro expansion failed for%") |
There was a problem hiding this comment.
Won't this logic cover any macro expansion failure warning, whether or not compile_error! or panic! is involved?
There was a problem hiding this comment.
I'll take a deeper look at this one...
| @@ -8,6 +8,9 @@ private import PathResolution | |||
| /** Holds if `p` may resolve to multiple items including `i`. */ | |||
| query predicate multiplePathResolutions(Path p, ItemNode i) { | |||
| p.fromSource() and | |||
| not p.(AstNode).isInMacroExpansion() and | |||
| // exclude paths inside attributes (e.g. `#[tokio::main]`) | |||
| not p.getParentNode*() instanceof Attr and | |||
There was a problem hiding this comment.
Why do we exclude paths inside attributes?
There was a problem hiding this comment.
good point, that does seem like sweeping a problem under the rug
| encoding = "0.2" | ||
| lazy_static = "1.5.0" | ||
| serde_json = "1.0.145" | ||
| serde_json = "1.0.149" |
There was a problem hiding this comment.
Why do we change files in ruby and shared? Do they have to all reference the same versions?
There was a problem hiding this comment.
The default instruction go for a generic update of everything using the Cargo.toml workspace at the root, which includes ruby/shared. We could probably limit to just rust-analyzer, or at least separate rust-analyzer from everything else
There was a problem hiding this comment.
Do you think this one is worth doing in this pull request?
Adds upgrade.ql with transformations for all schema changes from rust-analyzer 0.0.301 to 0.0.328: - Meta split into PathMeta/KeyValueMeta/TokenTreeMeta/UnsafeMeta - TraitAlias merged into Trait - BlockExpr.isTry() -> TryBlockModifier - StructField.getDefault() -> getDefaultVal() returning ConstArg - Variant.getDiscriminant() -> getConstArg() returning ConstArg - FormatArgsArg.getName() -> getArgName() returning FormatArgsArgName Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
WIP. Let's see what the CI thinks of this. (For the avoidance of doubt, Copilot wrote all of the code, and I'm still working my way through the changes themselves.)