]> git.proxmox.com Git - cargo.git/log
cargo.git
3 years agoAuto merge of #9202 - gauntface:patch-1, r=ehuss
bors [Mon, 1 Mar 2021 21:30:39 +0000 (21:30 +0000)]
Auto merge of #9202 - gauntface:patch-1, r=ehuss

Correct default cargo new edition

I think this is a small typo in the docs but feel free to close this PR if I'm mistaken

3 years agoAuto merge of #9224 - ehuss:split-debuginfo-docs, r=Eh2406
bors [Mon, 1 Mar 2021 20:43:44 +0000 (20:43 +0000)]
Auto merge of #9224 - ehuss:split-debuginfo-docs, r=Eh2406

Update split-debuginfo docs around the default.

The default behavior was changed during the course PR #9112, but the docs weren't updated.

Closes #9218

3 years agoUpdate split-debuginfo docs around the default.
Eric Huss [Mon, 1 Mar 2021 20:33:38 +0000 (12:33 -0800)]
Update split-debuginfo docs around the default.

3 years agoAuto merge of #9213 - ehuss:registry-api-errors, r=alexcrichton
bors [Mon, 1 Mar 2021 15:13:36 +0000 (15:13 +0000)]
Auto merge of #9213 - ehuss:registry-api-errors, r=alexcrichton

Minor update to registry API error messages.

This is a minor update to the registry API errors, trying to make them a little clearer and more helpful. My concerns were:

* `api errors (status 200 OK): some error message` — why was there both an error and 200 OK?
* `api errors` — What is an "api error" anyways?  The user is probably not familiar with the "registry API".
* Adds the URL of the server it is actually talking to, in case you may be confused when using a registry other than crates.io. This also tries to make it clearer that the message is coming from the remote server, and not cargo itself.

3 years agoAuto merge of #9214 - ehuss:clippy-cleanup, r=Eh2406
bors [Mon, 1 Mar 2021 04:25:06 +0000 (04:25 +0000)]
Auto merge of #9214 - ehuss:clippy-cleanup, r=Eh2406

Some minor code cleanup.

Some cleanup as recommended by clippy.  I think most of these are a little cleaner, though sometimes it is hard call.

3 years agoSome minor code cleanup.
Eric Huss [Mon, 1 Mar 2021 03:03:06 +0000 (19:03 -0800)]
Some minor code cleanup.

3 years agoMinor update to registry API error messages.
Eric Huss [Sat, 27 Feb 2021 20:38:17 +0000 (12:38 -0800)]
Minor update to registry API error messages.

3 years agoAuto merge of #9212 - vidbina:fix-spelling-worksapce-to-workspace, r=ehuss
bors [Sat, 27 Feb 2021 17:29:42 +0000 (17:29 +0000)]
Auto merge of #9212 - vidbina:fix-spelling-worksapce-to-workspace, r=ehuss

doc: Fix spelling worksapce->workspace

only in comments or documentation

3 years agodoc: Fix spelling worksapce->workspace
David Asabina [Sat, 27 Feb 2021 16:47:37 +0000 (17:47 +0100)]
doc: Fix spelling worksapce->workspace

only in comments or documentation

3 years agoAuto merge of #9209 - ehuss:update-spdx, r=alexcrichton
bors [Fri, 26 Feb 2021 15:28:38 +0000 (15:28 +0000)]
Auto merge of #9209 - ehuss:update-spdx, r=alexcrichton

Update SPDX version in docs.

The version is being updated in https://github.com/rust-lang/crates.io/pull/3335.

3 years agoUpdate SPDX version in docs.
Eric Huss [Fri, 26 Feb 2021 01:25:34 +0000 (17:25 -0800)]
Update SPDX version in docs.

3 years agoAuto merge of #8939 - Andy-Python-Programmer:master, r=alexcrichton
bors [Thu, 25 Feb 2021 19:40:08 +0000 (19:40 +0000)]
Auto merge of #8939 - Andy-Python-Programmer:master, r=alexcrichton

Throw error if CARGO_TARGET_DIR is an empty string

This pull request makes the target dir to be target/ if `CARGO_TARGET_DIR` is `` with spaces trimmed and not delete the current project.

Fixes: #8866
3 years agoAuto merge of #9207 - ehuss:testsuite-splitdebug, r=Eh2406
bors [Thu, 25 Feb 2021 19:06:49 +0000 (19:06 +0000)]
Auto merge of #9207 - ehuss:testsuite-splitdebug, r=Eh2406

testsuite: Use split debuginfo on macos.

This switches the testsuite to use "unpacked" debuginfo on macos, which is a substantial performance boost. On my system, the testsuite runs 1.55 times faster with this change.  Along with #9206, total testsuite time is 3.1 times faster.

3 years agoAuto merge of #9206 - ehuss:testsuite-rustup, r=Eh2406
bors [Thu, 25 Feb 2021 17:24:58 +0000 (17:24 +0000)]
Auto merge of #9206 - ehuss:testsuite-rustup, r=Eh2406

testsuite: Improve performance when using rustup.

The rustup wrapper adds a considerable amount of overhead when running processes like `rustc`.  This overrides `PATH` when running the testsuite to prioritize the actual executables to avoid the wrapper.

I'm not 100% confident this won't break something somewhere, but it seems like it should be safe.

In my tests, this makes the testsuite 1.5 to 1.7 times faster.

3 years agotestsuite: Improve performance when using rustup.
Eric Huss [Thu, 25 Feb 2021 16:17:05 +0000 (08:17 -0800)]
testsuite: Improve performance when using rustup.

3 years agotestsuite: Use split debuginfo on macos.
Eric Huss [Thu, 25 Feb 2021 17:06:11 +0000 (09:06 -0800)]
testsuite: Use split debuginfo on macos.

3 years agoFix tests
unknown [Thu, 25 Feb 2021 07:48:05 +0000 (18:48 +1100)]
Fix tests

3 years agoFixes error message
unknown [Thu, 25 Feb 2021 07:10:59 +0000 (18:10 +1100)]
Fixes error message

3 years agoUpdate manifest.md
Matt Gaunt [Wed, 24 Feb 2021 22:50:09 +0000 (14:50 -0800)]
Update manifest.md

Based on feedback, removed invalid value statement.

3 years agoAuto merge of #9128 - poliorcetics:respect-shortness-rustdoc, r=ehuss
bors [Wed, 24 Feb 2021 16:51:20 +0000 (16:51 +0000)]
Auto merge of #9128 - poliorcetics:respect-shortness-rustdoc, r=ehuss

Pass the error message format to rustdoc

- Goes with rust-lang/rust#81675.
- Will help with rust-lang/rust#81662.

This is my first PR to Cargo and I haven't finished reading the contributor guide yet, how should I add tests for this ? Did I had the code in the correct place ?

3 years agoAuto merge of #9203 - ehuss:fix-target_in_environment_contains_lower_case, r=alexcrichton
bors [Wed, 24 Feb 2021 15:19:55 +0000 (15:19 +0000)]
Auto merge of #9203 - ehuss:fix-target_in_environment_contains_lower_case, r=alexcrichton

Fix test target_in_environment_contains_lower_case

This test will fail if you actually have the x86_64-unknown-linux-musl target installed.  Instead of assuming it will fail, this uses the host target instead, which should only fail on windows due to case-insensitive environment keys.

3 years agoAdd tests
unknown [Wed, 24 Feb 2021 06:56:20 +0000 (17:56 +1100)]
Add tests

3 years agoAuto merge of #9201 - ehuss:close-error-during-drain, r=Eh2406
bors [Wed, 24 Feb 2021 04:40:06 +0000 (04:40 +0000)]
Auto merge of #9201 - ehuss:close-error-during-drain, r=Eh2406

Fix hang on broken stderr.

If stderr is closed and cargo tries to print a status message such as "Compiling", cargo can get into a hung state. This is because the `DrainState::run` function would insert the job into the `active` queue, and then return an error without starting the job. Since the job isn't started, there is nothing to remove it from the `active` queue, and thus cargo hangs forever waiting for it to finish.

The solution is to move the call to `note_working_on` earlier before the job is placed into the active queue.

This addresses the issue noted in https://github.com/rust-lang/cargo/issues/8714#issuecomment-783787163.

3 years agoUpdate manifest.md
Matt Gaunt [Wed, 24 Feb 2021 04:18:12 +0000 (20:18 -0800)]
Update manifest.md

Another alterations after pondering further

3 years agoUpdate manifest.md
Matt Gaunt [Wed, 24 Feb 2021 04:13:24 +0000 (20:13 -0800)]
Update manifest.md

Suggestion for text when the edition isn't defined at all

3 years agoAuto merge of #9195 - guswynn:test_name, r=ehuss
bors [Wed, 24 Feb 2021 03:59:28 +0000 (03:59 +0000)]
Auto merge of #9195 - guswynn:test_name, r=ehuss

Make it more clear which module is being tested when running cargo test

I recently asked in zulip if this is a good idea, as I find it hard to find the module thats being tested from the complex path with the hash.

Output of `cargo test`:
```
2021-02-21 13:29:33 I > ~/repos/cargo/target/debug/cargo test
    Finished test [unoptimized + debuginfo] target(s) in 0.42s
     Running unittests (target/debug/deps/test_tests-759130ea61f71571)

running 1 test
test tests::unit_test ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in
0.00s

     Running tests2/lib.rs (target/debug/deps/integration_tests-6b7f9fcd1721f083)

running 2 tests
test tests2_lib ... ok
test second_test::tests2_second ... ok

test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in
0.00s

     Running tests/lib.rs (target/debug/deps/lib-d2be6d29597c2790)

running 2 tests
test second_test::tests_i_am_run_twice ... ok
test tests_lib ... ok

test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in
0.00s
...
```

3 years agoCorrect default cargo new edition
Matt Gaunt [Wed, 24 Feb 2021 03:57:03 +0000 (19:57 -0800)]
Correct default cargo new edition

I think this is a small typo in the docs but feel free to close this PR if I'm mistaken

3 years agoFix test target_in_environment_contains_lower_case
Eric Huss [Wed, 24 Feb 2021 03:51:42 +0000 (19:51 -0800)]
Fix test target_in_environment_contains_lower_case

3 years agoFix hang on broken stderr.
Eric Huss [Wed, 24 Feb 2021 02:15:44 +0000 (18:15 -0800)]
Fix hang on broken stderr.

3 years agoAuto merge of #9184 - ehuss:fix-next-edition, r=alexcrichton
bors [Tue, 23 Feb 2021 21:35:42 +0000 (21:35 +0000)]
Auto merge of #9184 - ehuss:fix-next-edition, r=alexcrichton

Updates to edition handling.

This introduces some updates for edition handling (split into commits for review).  In short:

* `cargo-features = ["edition2021"]` can be used to opt-in to 2021 support without needing to pass around `-Z unstable-options`. I tried to emphasize in the docs that this is only for testing and experimentation.
* Make `"2"` the default resolver for 2021 edition.
* Make `cargo fix --edition` mean the "next" edition from the present one, and support 2021. Also, if already at the latest edition, generate a warning instead an error.
* I decided to allow `cargo fix --edition` from 2018 to 2021 on the nightly channel without an explicit opt-in. It's tricky to implement with an opt-in (see comment in diff).

Partial for #9048.
Fixes #9047.

3 years agoAuto merge of #9002 - volks73:feature-rustc-cfg-argument, r=alexcrichton
bors [Tue, 23 Feb 2021 17:20:26 +0000 (17:20 +0000)]
Auto merge of #9002 - volks73:feature-rustc-cfg-argument, r=alexcrichton

Add --cfg and --rustc-cfg flags to output compiler configuration

This PR is my attempt to address #8923.

I have added the `--cfg` flag to the `cargo rustc` subcommand and the `--rustc-cfg` flag to the `cargo build`, `cargo check`, `cargo test`, and `cargo bench` subcommands, respectively. Both versions of the flag, `--cfg` and `--rustc-cfg`, do the same thing, but I thought it looked weird for the `cargo rustc` subcommand to be `cargo rustc --rustc-cfg`. The following example invocations are possible, once stabilized:

- `cargo rustc --cfg`
- `cargo build --rustc-cfg`
- `cargo check --rustc-cfg`
- `cargo test --rustc-cfg`
- `cargo bench --rustc-cfg`

In each case, compilation is aborted and only compiler configuration is emitted. All of the context creation and configuration is still executed, but execution of the compilation job is aborted. Similar to the `--unit-graph` implementation, the `--cfg/--rustc-cfg` flag is hidden, marked as a "unstable", and requires the `-Z unstable-options` flag at the moment. A complete example invocation with this PR would be:

```bash
$ cargo +nightly rustc -Z unstable-options --cfg
```

I am open to alternatives for the flag name. I have thought of `--compiler-cfg`, `--compile-cfg`, and `--target-cfg`, but I went with `--rustc-cfg` because it is the Rust compiler (rustc) configuration (cfg). The `--target-cfg` could be confusing because there are Cargo targets and Compiler targets. A lone `--cfg` for the build, check, test, and bench subcommands would also be ambiguous and configuration that is being displayed.

Originally, I was only going to add the `--cfg` flag to the `cargo rustc` subcommand, but build, check, test, and bench all have the `--unit-graph` flag, and I used that flag's implementation and existence as a template for this implementation. I am not opposed to having just the `--cfg` flag for the `cargo rustc` subcommand as originally proposed in #8923.

I discovered during my initial investigation to implement the feature and familiarization with the Cargo codebase, that as part of the build context creation and compilation process, Cargo internally calls the `rustc --print cfg` command for all targets and stores the output in the `RustcTargetData` type. It does this for the host and any explicitly defined targets for cross-compilation. Compilation features, such as `+crt-static`, added to the compilation process through Cargo environment variables or TOML configuration directives are added to the internal `rustc --print cfg` command call. So, the `--cfg/--rustc-cfg` just emits the contents of the `RustcTagetData` type as JSON. There is no need to call the `rustc --print cfg` command again. The implementation then becomes nearly identical to the `--unit-graph` implementation.

The output is only in JSON, similar to the `--unit-graph` feature, instead of the key-value and name style of the `rustc --print cfg` output because it easily resolves issues related to multi-targets, explicit target (`--target <TRIPLE>`) versus host configurations, and cross compilation. Subcommands and other projects can parse the JSON to recreate the key-value and name style if desired based on a specific target or the host. Here is an example of the JSON output (formatted for humans, the actual output is condensed), which is also available as a comment in the `cargo::core::compiler::rustc_cfg` module:

```javascript
{
     "version": 1,
     "host": {
         "names": ["windows", "debug_assertions"],
         "arch":"x86_64",
         "endian":"little",
         "env":"msvc",
         "family":"windows",
         "features":["fxsr","sse","sse2"],
         "os":"windows",
         "pointer_width":"64",
         "vendor":"pc"
     },
     "targets": {
         "x86_64-unknown-linux-gnu": {
             "names": ["debug_assertions", "unix"],
             "arch":"x86_64",
             "endian":"little",
             "env":"gnu",
             "family":"unix",
             "features": ["fxsr","sse","sse2"],
             "os":"linux",
             "pointer_width":"64",
             "vendor":"unknown"
         },
         "i686-pc-windows-msvc": {
             "names": ["windows", "debug_assertions"],
             "arch":"x86",
             "endian":"little",
             "env":"msvc",
             "family":"windows",
             "features":["fxsr","sse","sse2"],
             "os":"windows",
             "pointer_width":"32",
             "vendor":"pc"
         }
     }
 }
```

I decided to remove the "target_" prefix from the relevant configurations to reduce "noise" in the output. Again, I am open to alternatives or suggestions on the JSON format.

3 years agoAuto merge of #9105 - Swatinem:rustdoc-run-cwd, r=alexcrichton
bors [Tue, 23 Feb 2021 16:39:45 +0000 (16:39 +0000)]
Auto merge of #9105 - Swatinem:rustdoc-run-cwd, r=alexcrichton

Run rustdoc doctests relative to the workspace

By doing so, rustdoc will also emit workspace-relative filenames for the doctests.

This was first landed in #8954 but later backed out in #8996 because it changed the CWD of rustdoc test invocations.

The second try relies on the new `--test-run-directory` rustdoc option which was added in https://github.com/rust-lang/rust/pull/81264 to explicitly control the rustdoc test cwd.

fixes #8993

3 years agoAuto merge of #9175 - wickerwaka:env-section, r=alexcrichton
bors [Tue, 23 Feb 2021 15:22:59 +0000 (15:22 +0000)]
Auto merge of #9175 - wickerwaka:env-section, r=alexcrichton

Add support for [env] section in .cargo/config.toml

This adds support for an `[env]` section in the config.toml files. Environment variables set in this section will be applied to the environment of any processes executed by cargo.

This is implemented to follow the recommendations in https://github.com/rust-lang/cargo/pull/8839#issuecomment-725678657

Variables have optional `force` and `relative` flags. `force` means the variable can override an existing environment variable. `relative` means the variable represents a path relative to the location of the directory that contains the `.cargo/` directory that contains the `config.toml` file. A relative variable will have an absolute path prepended to it before setting it in the environment.

```
[env]
FOOBAR = "Apple"
PATH_TO_SOME_TOOL = { value = "bin/tool", relative = true }
USERNAME = { value = "test_user", force = true }
```

Fixes #4121

3 years agoRun rustdoc doctests relative to the workspace
Arpad Borsos [Mon, 22 Feb 2021 19:06:03 +0000 (20:06 +0100)]
Run rustdoc doctests relative to the workspace

By doing so, rustdoc will also emit workspace-relative filenames for the doctests.

This was first landed in #8954 but later backed out in #8996 because it changed the CWD of rustdoc test invocations.

The second try relies on the new `--test-run-directory` rustdoc option which was added in https://github.com/rust-lang/rust/pull/81264 to explicitly control the rustdoc test cwd.

fixes #8993

3 years agoAuto merge of #9161 - ehuss:index-v-features2, r=alexcrichton
bors [Mon, 22 Feb 2021 15:48:02 +0000 (15:48 +0000)]
Auto merge of #9161 - ehuss:index-v-features2, r=alexcrichton

Add schema field and `features2` to the index.

This adds a `v` field to the index which indicates a format version for an index entry. If Cargo encounters a version newer than it understands, it will ignore those entries. This makes it safer to make changes to the index entries (such as adding new things), and not need to worry about how older cargos will react to it. In particular, this will make it safer to run `cargo update` on older versions if we ever decide to add new things to the index.

Currently this is not written anywhere, and is intended as a safety guard for the future. For now I will leave it undocumented until we actually decide to start using it.

This also moves the new syntax for namespaced features and weak dependency features into a new field ("features2") in the index. This is necessary to avoid breaking Cargo versions older than 1.19, which fail to parse the index even if there is a Cargo.lock file.

It is intended that only crates.io will bother with creating this field. Other registries don't need to bother, since they generally don't support Cargo older than 1.19.

I'm uncertain exactly when we should try to update crates.io to start accepting this, as that is a somewhat permanent decision.

3 years agoFix usage of assert methods
Chris Field [Mon, 22 Feb 2021 03:58:31 +0000 (22:58 -0500)]
Fix usage of assert methods

The `with_stdout_contains` was mis-used. Since some lines may or may not
appear for some compiler targets and environments (nightly, beta,
stable, etc.) the tests would fail because the output was not identical.
Instead of a using raw strings, each line with the arch, endian, env,
family, vendor, pointer_width, etc. that are known to always be
present (at least of the CI builds) are included. This should work for
the CI environments and my local environment.

3 years agoFix tests for CI environment
Chris Field [Mon, 22 Feb 2021 02:45:55 +0000 (21:45 -0500)]
Fix tests for CI environment

The `panic="unwind"` appears in the output for the CI tests, but not in
my local tests. I need to investigate the origin of this configuration
but it causes the CI builds to fail.

3 years agoFix formatting
Chris Field [Mon, 22 Feb 2021 01:00:28 +0000 (20:00 -0500)]
Fix formatting

3 years agoAdd test with cargo configuration file
Chris Field [Mon, 22 Feb 2021 00:58:19 +0000 (19:58 -0500)]
Add test with cargo configuration file

A `.cargo/config.toml` file is used to add the "crt-static" target
feature and test printing the compiler target configuration contains the
`target_feature="crt-static"` line.

3 years agoAdd test using RUSTFLAGS env var
Chris Field [Mon, 22 Feb 2021 00:54:08 +0000 (19:54 -0500)]
Add test using RUSTFLAGS env var

The `RUSTFLAGS` environment variable is used to add the "crt-static"
target feature and test printing the target compiler configuration
contains the `target_feature="crt-static"` line.

3 years agoAdd multitarget test
Chris Field [Mon, 22 Feb 2021 00:50:21 +0000 (19:50 -0500)]
Add multitarget test

Using the multitarget feature to print the configuration of multiple
compiler target configurations, a test is created.

3 years agoAdd first test
Chris Field [Mon, 22 Feb 2021 00:35:33 +0000 (19:35 -0500)]
Add first test

Not sure how to handle running the test on different hosts, so the first
test is explicit about the compile target to print the configuration.

3 years agofix test suite
Gus Wynn [Sun, 21 Feb 2021 22:37:42 +0000 (14:37 -0800)]
fix test suite

3 years agoMake it more clear which module is being tested when running cargo test
Gus Wynn [Sun, 21 Feb 2021 21:35:25 +0000 (13:35 -0800)]
Make it more clear which module is being tested when running cargo test

3 years agoChange env section parsing
Martin Donlon [Sat, 20 Feb 2021 21:49:08 +0000 (13:49 -0800)]
Change env section parsing

`untagged` enum variants are not recognised by serde if the variant contains a `Value<T>` type.
This change uses an transparent "inner" enum `Value<T>` member, which is handled correctly by serde.

3 years agoAuto merge of #9189 - weihanglo:issue-9137, r=ehuss
bors [Sat, 20 Feb 2021 15:45:19 +0000 (15:45 +0000)]
Auto merge of #9189 - weihanglo:issue-9137, r=ehuss

Document the default location where cargo install emitting build artifacts

Resolves #9137

Document the default location where `cargo install` emitting build artifacts.

3 years agoFix warning with unused import
Chris Field [Sat, 20 Feb 2021 15:40:08 +0000 (10:40 -0500)]
Fix warning with unused import

3 years agoChange to forwarding rustc stdio to cargo
Chris Field [Sat, 20 Feb 2021 15:35:27 +0000 (10:35 -0500)]
Change to forwarding rustc stdio to cargo

The `exec_with_output` function and subsequent writing to stdout and
stderr is replaced with the `exec` function that inherits the stdio
streams from Cargo is used.

Weirdly, I originally tried to use the `exec` function but I go no
output. The `cargo run -- rustc --print cfg` appear to just hang. So, I
switched to using the `exec_with_output` function. Now, the `exec`
function appears to work and it is much simplier. There must have been
something else going on with my system at the time.

3 years agoFix index_v version check.
Eric Huss [Sat, 20 Feb 2021 14:28:51 +0000 (06:28 -0800)]
Fix index_v version check.

The whole point is that when loading a cache from an earlier version,
that cache may be missing new entries. So don't load older caches.

3 years agodoc: generated by build-man.sh
Weihang Lo [Sat, 20 Feb 2021 03:33:23 +0000 (11:33 +0800)]
doc: generated by build-man.sh

3 years agodoc(man): clarify `--path` influnce on `target-dir`
Weihang Lo [Sat, 20 Feb 2021 03:31:09 +0000 (11:31 +0800)]
doc(man): clarify `--path` influnce on `target-dir`

3 years agoChange how workspace resolve behavior is calculated.
Eric Huss [Fri, 19 Feb 2021 22:23:39 +0000 (14:23 -0800)]
Change how workspace resolve behavior is calculated.

This fixes an issue where warnings would be printed for packages
migrating to 2021 in a workspace (that the "resolver" field is ignored,
which is wrong).

This also places the default resolver logic in one place, and should
make it easier to update later.

3 years ago`cargo fix`: Remove --prepare-for option.
Eric Huss [Fri, 19 Feb 2021 21:08:59 +0000 (13:08 -0800)]
`cargo fix`: Remove --prepare-for option.

This was deprecated, never officially part of the stable release.

3 years agoAdd some more details to the `cargo fix --edition` procedure.
Eric Huss [Fri, 19 Feb 2021 20:58:30 +0000 (12:58 -0800)]
Add some more details to the `cargo fix --edition` procedure.

3 years agoMake edition transitions easier.
Eric Huss [Fri, 19 Feb 2021 20:57:36 +0000 (12:57 -0800)]
Make edition transitions easier.

This attempts to centralize all the edition stuff in one place (the
`Edition` enum) so that adding a new edition or stabilizing one should
be relatively little work (and more importantly, we don't miss things).

Importantly, this changes `cargo new` to default to the latest stable.
It also changes the `cargo fix --edition-idiom` behavior to only apply
idioms for the *current* edition.

3 years agoInclude the index version in the index cache.
Eric Huss [Fri, 19 Feb 2021 18:51:38 +0000 (10:51 -0800)]
Include the index version in the index cache.

This is intended to help prevent the following scenario from happening:

1. Old cargo builds an index cache.
2. Old cargo finds an index entry it cannot parse, skipping it,
   and saving the cache without the entry.
3. New cargo loads the cache with the missing entry, and never sees
   the new entries that it understands.

This may result in more cache thrashing, but that seems better than
having new cargos missing entries.

3 years agoAuto merge of #9185 - horacimacias:master, r=ehuss
bors [Fri, 19 Feb 2021 15:32:24 +0000 (15:32 +0000)]
Auto merge of #9185 - horacimacias:master, r=ehuss

Do not exit prematurely if anything failed installing.

https://github.com/rust-lang/cargo/issues/9180

3 years agodoc: generated by build-man.sh
Weihang Lo [Fri, 19 Feb 2021 14:00:57 +0000 (22:00 +0800)]
doc: generated by build-man.sh

3 years agodoc(man): add temp-target-dir variable
Weihang Lo [Fri, 19 Feb 2021 14:00:15 +0000 (22:00 +0800)]
doc(man): add temp-target-dir variable

3 years agoDo not exit prematurely if anything failed installing.
Horaci Macias [Fri, 19 Feb 2021 08:53:46 +0000 (09:53 +0100)]
Do not exit prematurely if anything failed installing.
https://github.com/rust-lang/cargo/issues/9180

3 years agoAuto merge of #9182 - alexcrichton:fix-lto-off, r=ehuss
bors [Thu, 18 Feb 2021 15:49:14 +0000 (15:49 +0000)]
Auto merge of #9182 - alexcrichton:fix-lto-off, r=ehuss

Propagate `lto=off` harder

This commit fixes an issue with LTO calculation for various units when
`lto=off` is specified in the profile. This ensures now that `lto=off`
is passed to all transitive dependencies as well to disable thin-local
LTO. As an added bonus this also passed `embed-bitcode=no` whenever
`lto=off` is specified since we know we won't be using bitcode anyway.

Closes #9171

3 years agoAuto merge of #9183 - weihanglo:refactor/straighforward-deref, r=alexcrichton
bors [Thu, 18 Feb 2021 15:11:16 +0000 (15:11 +0000)]
Auto merge of #9183 - weihanglo:refactor/straighforward-deref, r=alexcrichton

refactor: make deref intentions more straightforward

Just some little tweaks to make intentions of deref more straightforward.

3 years agoDo not exit prematurely if anything failed installing.
Horaci Macias [Thu, 18 Feb 2021 08:41:17 +0000 (09:41 +0100)]
Do not exit prematurely if anything failed installing.
https://github.com/rust-lang/cargo/issues/9180

3 years agoUpdate `cargo fix --edition` docs.
Eric Huss [Thu, 18 Feb 2021 06:14:35 +0000 (22:14 -0800)]
Update `cargo fix --edition` docs.

# Conflicts:
# src/cargo/ops/fix.rs
# src/doc/src/reference/unstable.md

3 years agoUpdates to `cargo fix --edition`.
Eric Huss [Thu, 18 Feb 2021 06:12:46 +0000 (22:12 -0800)]
Updates to `cargo fix --edition`.

* `--edition` always means "next" edition.
* `--edition` when on the most recent edition is not an error, just a warning.
* Support fix to 2021 edition.

3 years agoMake "2" the default resolver in the 2021 edition.
Eric Huss [Thu, 18 Feb 2021 06:06:58 +0000 (22:06 -0800)]
Make "2" the default resolver in the 2021 edition.

3 years agoAdd edition2021 cargo feature.
Eric Huss [Thu, 18 Feb 2021 06:01:42 +0000 (22:01 -0800)]
Add edition2021 cargo feature.

This is intended to help make it easier to test the 2021 edition.

3 years agoAdd a migrating message for `cargo fix --edition`.
Eric Huss [Thu, 18 Feb 2021 05:36:28 +0000 (21:36 -0800)]
Add a migrating message for `cargo fix --edition`.

This helps indicate which edition you are moving from and to.

3 years agoChange Fixing to Fixed, and add a verbose "Fixing".
Eric Huss [Thu, 18 Feb 2021 04:41:38 +0000 (20:41 -0800)]
Change Fixing to Fixed, and add a verbose "Fixing".

What was previously "Fixing" was a message for after the fixes had
been applied. I think it would be clearer if it said "Fixed",
to indicate that the fixes had actually finished.

The new "Fixing" is posted just before it starts. This is verbose-only
since it is a little noisy.

3 years agorefactor: make deref intentions more clear
Weihang Lo [Thu, 18 Feb 2021 04:16:23 +0000 (12:16 +0800)]
refactor: make deref intentions more clear

3 years agoSome minor refactoring of `fix` code.
Eric Huss [Thu, 18 Feb 2021 04:11:15 +0000 (20:11 -0800)]
Some minor refactoring of `fix` code.

There shouldn't be any functional changes here.

* Some doc comments.
* Construct `FixArgs` at once so it doesn't need to bother with unnecessary Options.
* Remove IdiomEditionMismatch, it is not used.
* Use a general deduping mechanism for fix messages.

3 years agoPropagate `lto=off` harder
Alex Crichton [Wed, 17 Feb 2021 19:40:38 +0000 (11:40 -0800)]
Propagate `lto=off` harder

This commit fixes an issue with LTO calculation for various units when
`lto=off` is specified in the profile. This ensures now that `lto=off`
is passed to all transitive dependencies as well to disable thin-local
LTO. As an added bonus this also passed `embed-bitcode=no` whenever
`lto=off` is specified since we know we won't be using bitcode anyway.

Closes #9171

3 years agoAuto merge of #9174 - ehuss:prelude-no_std, r=alexcrichton
bors [Tue, 16 Feb 2021 16:38:07 +0000 (16:38 +0000)]
Auto merge of #9174 - ehuss:prelude-no_std, r=alexcrichton

Update link for no_std attribute.

There was some reorganization in the reference as part of https://github.com/rust-lang/reference/pull/937.

3 years agoAdd `-Z configurable-env` to gate [env] usage
Martin Donlon [Tue, 16 Feb 2021 04:21:22 +0000 (20:21 -0800)]
Add `-Z configurable-env` to gate [env] usage

Added the `-Z configurable-env` option which controls whether the data from the [env] section is used. Updated the tests to pass the flag and to masquerade as the nightly cargo.

3 years agoRun `cargo fmt --all`
Martin Donlon [Mon, 15 Feb 2021 23:29:08 +0000 (15:29 -0800)]
Run `cargo fmt --all`

Fix formatting errors

3 years agoAdd support for [env] section in .cargo/config.toml
Martin Donlon [Mon, 15 Feb 2021 23:17:00 +0000 (15:17 -0800)]
Add support for [env] section in .cargo/config.toml

This adds support for an `[env]` section in the config.toml files. Environment variables set in this section will be applied to the environment of any processes executed by cargo.

```
[env]
FOOBAR = "Apple"
PATH_TO_SOME_TOOL = { value = "bin/tool", relative = true }
USERNAME = { value = "test_user", force = true }
```

3 years agoAuto merge of #9173 - casey:patch-2, r=ehuss
bors [Mon, 15 Feb 2021 18:01:39 +0000 (18:01 +0000)]
Auto merge of #9173 - casey:patch-2, r=ehuss

Remove mention of --message-format taking multiple values

I tried passing multiple values to `--message-format`, as is mentioned in the docs, and it didn't work:

```
$ cargo test --message-format human,json
error: cannot specify two kinds of `message-format` arguments
$ cargo test --message-format human --message-format json
error: cannot specify two kinds of `message-format` arguments
````

It's entirely possible I misunderstood the docs, but just in case I didn't, I thought I'd open this PR to remove the bit that mentions that the flag accepts multiple values.

3 years agoUpdate link for no_std attribute.
Eric Huss [Mon, 15 Feb 2021 03:35:11 +0000 (19:35 -0800)]
Update link for no_std attribute.

3 years agoClarify which values may be used together with `--message-format`
Casey Rodarmor [Mon, 15 Feb 2021 01:58:26 +0000 (17:58 -0800)]
Clarify which values may be used together with `--message-format`

3 years agoAuto merge of #9169 - Liberatys:emit-warning-on-env-variable-case-mismatch, r=ehuss
bors [Sun, 14 Feb 2021 19:35:20 +0000 (19:35 +0000)]
Auto merge of #9169 - Liberatys:emit-warning-on-env-variable-case-mismatch, r=ehuss

Emit warning on env variable case mismatch

When running a command like `cargo --target TRIPPLE` cargo expects to find the environment variable CARGO_TARGET_[TRIPPLE]_* with uppercase and underscores. This check emits a warning if the checked environment variable has a mismatching case and/or contains dashes rather than underscores. The warning contains the given env variable as well as an explanation for the cause of the warning.

The check is skipped on windows as environment variables are treated as case insensitive on the platform.

Fixes #8285

3 years agoCombine target env tests.
Eric Huss [Sun, 14 Feb 2021 19:32:18 +0000 (11:32 -0800)]
Combine target env tests.

3 years agoTest the passing of --error-format to rustdoc
Alexis Bourget [Sun, 14 Feb 2021 17:01:50 +0000 (18:01 +0100)]
Test the passing of --error-format to rustdoc

3 years agoRe-add backticks
Nick Flueckiger [Sun, 14 Feb 2021 12:15:35 +0000 (13:15 +0100)]
Re-add backticks

3 years agoUpdate wording and use if let
Nick Flueckiger [Sun, 14 Feb 2021 12:00:20 +0000 (13:00 +0100)]
Update wording and use if let

3 years agoFix windows test case
Nick Flueckiger [Sat, 13 Feb 2021 23:20:21 +0000 (00:20 +0100)]
Fix windows test case

3 years agoUpdate windows check
Nick Flueckiger [Sat, 13 Feb 2021 22:20:43 +0000 (23:20 +0100)]
Update windows check

3 years agoAuto merge of #8640 - CPerezz:doc_versioning, r=ehuss
bors [Sat, 13 Feb 2021 21:57:02 +0000 (21:57 +0000)]
Auto merge of #8640 - CPerezz:doc_versioning, r=ehuss

Implement Rustdoc versioning checks

Before compiling, we need to make sure that if there were any previous docs already compiled, they were compiled with the same Rustc version that we're currently using. Otherways we must remove the `doc/` folder and compile again.

This is important because as stated in #8461 the .js/.html&.css files that are generated by Rustc don't have any versioning. Therefore, we can fall in weird bugs and behaviours if we mix different compiler versions of these js/.html&.css files.

Closes #8461

3 years agoUpdate src/cargo/core/compiler/build_context/target_info.rs
Carlos Pérez [Sat, 13 Feb 2021 20:40:09 +0000 (21:40 +0100)]
Update src/cargo/core/compiler/build_context/target_info.rs

Co-authored-by: Eric Huss <eric@huss.org>
3 years agoAdapt implementation with reviewer suggestions and requested changes
Nick Flueckiger [Sat, 13 Feb 2021 20:17:18 +0000 (21:17 +0100)]
Adapt implementation with reviewer suggestions and requested changes

3 years agoCorrect target
Nick Flueckiger [Sat, 13 Feb 2021 11:19:00 +0000 (12:19 +0100)]
Correct target

3 years agoLint
Nick Flueckiger [Sat, 13 Feb 2021 11:14:24 +0000 (12:14 +0100)]
Lint

3 years agoExclude test on windows
Nick Flueckiger [Sat, 13 Feb 2021 10:59:38 +0000 (11:59 +0100)]
Exclude test on windows

3 years agoImplement uppercase env map for key case mismatch
Nick Flueckiger [Sat, 13 Feb 2021 10:46:13 +0000 (11:46 +0100)]
Implement uppercase env map for key case mismatch

3 years agoAuto merge of #9168 - ehuss:version-bump, r=Eh2406
bors [Fri, 12 Feb 2021 21:12:02 +0000 (21:12 +0000)]
Auto merge of #9168 - ehuss:version-bump, r=Eh2406

Bump to 0.53.0, update changelog

3 years agoUpdate changelog for 1.51
Eric Huss [Fri, 12 Feb 2021 19:48:56 +0000 (11:48 -0800)]
Update changelog for 1.51

3 years agoBump to 0.53.0
Eric Huss [Fri, 12 Feb 2021 18:58:38 +0000 (10:58 -0800)]
Bump to 0.53.0

3 years agoAuto merge of #9164 - ehuss:config-search-stop, r=Eh2406
bors [Thu, 11 Feb 2021 14:48:46 +0000 (14:48 +0000)]
Auto merge of #9164 - ehuss:config-search-stop, r=Eh2406

Prevent testsuite from loading config out of sandbox.

This adds a limit to prevent config loading from walking outside of the test sandbox root. There was an environment variable for this, but that doesn't work too well for tests that were loading the config directly.

Fixes #9107

3 years agoPrevent testsuite from loading config out of sandbox.
Eric Huss [Thu, 11 Feb 2021 01:39:38 +0000 (17:39 -0800)]
Prevent testsuite from loading config out of sandbox.

3 years agoChange to use TargetInfo type
Chris Field [Thu, 11 Feb 2021 00:30:21 +0000 (19:30 -0500)]
Change to use TargetInfo type

The `TargetInfo` type, but more specifically the `TargetInfo::new`
method is used instead of the making the `env_args` function public and
using it.

This means the `env_args` function can go back to being private, too.

3 years agoChange scope of function
Chris Field [Thu, 11 Feb 2021 00:10:14 +0000 (19:10 -0500)]
Change scope of function

The `output_err_info` function can be private now that the
implementation for the `cargo rustc --print <FOO>` command has changed
to not capture the output.