cargo/init: avoid target.name assignments if possible
Make `cargo init` skip `target.name` assignments if the default value is suitable.
Currently, all paths of `cargo init` will set target-names to the package-name, ensuring that a suitable target-name is set. This is required for all targets but libraries. Unfortunately, library-names have more restrictions than other targets. For example, creating a library with dashes will lead to errors:
mkdir foo-bar
cd foo-bar
touch foo-bar.rs
cargo init --lib
Fortunately, target-names for libraries are inferred from the package-name. Hence, by omitting the target-name, a valid configuration will be produced.
Instead of adjusting `SourceFileInformation::target_name` to use `Option<String>`, this commit strips the field completely, since all callers set it to the package-name.
Fixes: #11259
chore: Fix minor grammar nit in command-line help
This change fixes a *very* minor grammar mistake. "Require" used in this way is
typically followed by the infinitive. In addition, since "up-to-date" is used
an adjective now it should be hyphenated.
See https://www.merriam-webster.com/dictionary/up-to-date
### Testing
I have run `cargo test` with `CFG_DISABLE_CROSS_TESTS=1` as well as `cargo
build-man` to regenerate the man pages.
Fixes#13601.
Make `cargo init` skip `target.name` assignments if the default value is
suitable.
Currently, all paths of `cargo init` will set target-names to the
package-name, ensuring that a suitable target-name is set. This is
required for all targets but libraries. Unfortunately, library-names
have more restrictions than other targets. For example, creating a
library with dashes will lead to errors:
mkdir foo-bar
cd foo-bar
touch foo-bar.rs
cargo init --lib
Fortunately, target-names for libraries are inferred from the
package-name. Hence, by omitting the target-name, a valid configuration
will be produced.
Instead of adjusting `SourceFileInformation::target_name` to use
`Option<String>`, this commit strips the field completely, since all
callers set it to the package-name.
Signed-off-by: David Rheinsberg <david@readahead.eu>
cargo: prevent dashes in lib.name
The TOML parser of Cargo currently refuses `lib.name` entries that contain dashes. Unfortunately, it uses the package-name as default if no explicit `lib.name` entry is specified. This package-name, however, can contain dashes.
Cargo documentation states that the package name is converted first, yet this was never implemented by the code-base.
Fix this inconsistency and convert the package name to a suitable crate-name first.
This fixes#12780. It is an alternative to #12640.
feat: Add 'open-namespaces' feature
### What does this PR try to resolve?
This is a step towards #13576
### How should we test and review this PR?
### Additional information
refactor: Expose source/spans to Manifest for emitting lints
### What does this PR try to resolve?
This is a follow up to #13589.
This does nothing on its own.
This is meant to short-circuit some of my refactorings so Muscraft can
start on their work on adding lints while I work to move out existing
warnings into being able to be converted to lints.
### How should we test and review this PR?
This includes documentation changes suggested in #13589
### Additional information
feat(tree): Control `--charset` via auto-detecting config value
### What does this PR try to resolve?
This tries to generalize `cargo tree --charset` so any part of cargo's output can use it. For example,
- `cargo search` could use this for fancier tables
- `cargo add` could use this for fancier feature lists
- #12235 could use this for fancy rendering like miette does (CC `@Muscraft` )
- Progress bars could use fancier characters <-- this is what I'm personally working towards
This builds on the idea from #12889 that we can use fancier terminal features so long as we have good auto-detection and provide users an escape hatch until the auto-detection is improved or in case they disagree.
As a side benefit
- `cargo tree` will auto-detect when the prior default of `--charset utf8` is a good choice
- Users can control `cargo tree --charset` via `.cargo/config.toml`
### How should we test and review this PR?
This is gradually introduced through the individual commits.
### Additional information
This does nothing on its own.
This is meant to short-circuit some of my refactorings so Muscraft can
start on their work on adding lints while I work to move out existing
warnings into being able to be converted to lints.
refactor(toml): Flatten manifest parsing
### What does this PR try to resolve?
This is just a clean up but the goals are
- Support diagnostics that show source by tracking `&str`, `ImDocument`, etc in `Manifest` by making each accessible in the creation of a `Manifest`
- Defer warning analysis until we know what is a local vs non-local workspace by refactoring warnings out into a dedicated step
- Centralize the logic for `cargo publish` stripping of dev-dependencies and their feature activations by allowing a `Summary` to be created from any "resolved" `TomlManifest`
- Enumerate all build targets in the "resolved" `TomlManifest` so they get included in `cargo publish`, reducing the work done on registry dependencies and resolving problems like #13456
Along the way, this fixed a bug where we were not reporting warnings from virtual manifests
### How should we test and review this PR?
### Additional information
Before, we split things up. This makes it so everything has access to
every step so we can reap the benefits
- use `&str` and `ImDocument` for diagnostics
- access `original`
fix: strip feature dep when dep is dev dep
### What does this PR try to resolve?
This change aims to strip features dependencies without a version key to be published.
If a dev-dependency is missing the version, it will be stripped from the packaged manifest.
The features table may contains the deps in following places.
- Target
- normal
- dev
- build
- normal-and-dev
- normal
- dev
- build
- normal-and-dev
### How should we test and review this PR?
See the initial commit, it shows current behavior that will cause error when feature has deps that point to dev_dep and doesn't have a version specified.
Title | orignal toml | published toml |
---- | ---- | ---- |
Before | feature = ["dev-dep/feature"] <br/> [dev-dependencies] <br/> dev-dep = { .., features: ["feature"] } | feature = ["dev-dep/feature"] <br/> [dev-dependencies] <br/> dev-dep = { .., features: ["feature"] } |
After | feature = ["dev-dep/feature"] <br/> [dev-dependencies] <br/> dev-dep = { .., features: ["feature"] } | feature = [] <br/> [dev-dependencies] ```
Fix: #12225
refactor(lockfile): Make diffing/printing more reusable
### What does this PR try to resolve?
This is prep for #13561 so we can tailor the printing of lockfile changes to each use without a bunch of flags.
### How should we test and review this PR?
### Additional information
test: Add tests for using worktrees and sparse checkouts
### What does this PR try to resolve?
Based on `@eminence's` [comment](https://github.com/rust-lang/cargo/issues/7876#issuecomment-1325974682), Add tests for using worktrees or spase checkouts.
### How should we test and review this PR?
Checkout and run tests:
```
cargo test --package cargo --test testsuite -- git::git_worktree_with_original_repo_renamed
```
```
cargo test --package cargo --test testsuite -- git::git_worktree_with_bare_original_repo
```
util/network/http: Use `cargo/1.2.3` user-agent header
... instead of `cargo 1.2.3`.
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/User-Agent#syntax declares that the product and product version are usually separated by a slash. This commit changes the cargo `User-Agent` header to follow that syntax instead of using whitespace for the separator.
