8812: fix: fix dependencies of build scripts r=jonas-schievink a=jonas-schievink
Previously, we added a dependency for all targets in a package to the package's library target. This is correct for most targets, except build scripts, which run before the library crate is built. This PR removes the incorrect dependency on the library target.
We also used to treat all dependencies the same, which led to build scripts being able to use regular dependencies as well as dev-dependencies. This is also fixed by this PR, and build scripts only depend on build-dependencies.
Incorrect dependency graph:
Fixed graph after this PR:
Co-authored-by: Jonas Schievink <jonasschievink@gmail.com>
8808: fix: Fix fn proc macro input again r=jonas-schievink a=jonas-schievink
https://github.com/rust-analyzer/rust-analyzer/pull/8806 broke the `TokenMap`, so none of the tokens in fn-like proc macro inputs could be related to the output (presumably this is because of the `clone_for_update`).
This PR instead just sets `delimiter = None;` after the `TokenMap` and `Subtree` are already created.
We should probably have more tests for fn-like proc macros, and consider making the behavior consistent with MBE (which *require* the delimiters to be present).
bors r+
Co-authored-by: Jonas Schievink <jonasschievink@gmail.com>
8806: fix: Strip delimiter from fn-like macro invocations r=jonas-schievink a=jonas-schievink
This broke in https://github.com/rust-analyzer/rust-analyzer/pull/8796 (again), the fix is easy though
bors r+
Co-authored-by: Jonas Schievink <jonasschievink@gmail.com>
8796: internal: rewrite `#[derive]` removal to be based on AST (take 2) r=jonas-schievink a=jonas-schievink
Second attempt of https://github.com/rust-analyzer/rust-analyzer/pull/8443, this uses syntactical attribute offsets in `hir_expand`, and changes `attr.rs` to make those easy to derive.
This will make it easy to add similar attribute removal for attribute macros, unblocking them.
Co-authored-by: Jonas Schievink <jonasschievink@gmail.com>
8800: feat: Make "pull assignments up" assist work in more cases r=Jesse-Bakker a=Jesse-Bakker
Fixes#8771
Co-authored-by: Jesse Bakker <github@jessebakker.com>
This recognizes `let a = [1u8, 2, 3]` as having type `[u8; 3]` instead
of the previous `[u8; _]`. Byte strings and `[0u8; 2]` kinds of range
array declarations are unsupported as before.
I don't know why a bunch of our rustc tests had single quotes inside
strings un-escaped by `UPDATE_EXPECT=1 cargo t`, but I don't think it's
bad? Maybe something in a nightly?
8794: Give MergeBehaviour variants better names r=Veykril a=Veykril
I never really liked the variant names I gave this enum from the beginning and then I found out about rustfmt's `imports_granularity` config:
> imports_granularity
>
> How imports should be grouped into use statements. Imports will be merged or split to the configured level of granularity.
>
> Default value: Preserve
> Possible values: Preserve, Crate, Module, Item
> Stable: No
I personally prefer using `crate` over `full` and `module` over last, they seem more descriptive. Keeping these similar between tooling also seems like a good plus point to me.
We might even wanna take over the entire enum at some point if we have a `format/cleanup imports` assists in the future which would probably want to also have the `preserve` and `item` options.
Co-authored-by: Lukas Wirth <lukastw97@gmail.com>
There's a tension between keeping a well-architectured minimal
orthogonal set of constructs, and providing convenience functions.
Relieve this pressure by introducing an dedicated module for
non-orthogonal shortcuts.
This is inspired by the django.shortcuts module which serves a similar
purpose architecturally.
