2024-02-19 17:09:47 +00:00
|
|
|
//! CI script used for Bevy.
|
|
|
|
|
2022-05-02 19:13:34 +00:00
|
|
|
use bitflags::bitflags;
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
use std::collections::BTreeMap;
|
|
|
|
use xshell::{cmd, Cmd, Shell};
|
2022-05-02 19:13:34 +00:00
|
|
|
|
|
|
|
bitflags! {
|
2024-04-10 14:29:05 +00:00
|
|
|
/// A collection of individual checks that can be run in CI.
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
|
2022-05-02 19:13:34 +00:00
|
|
|
struct Check: u32 {
|
2024-04-10 14:29:05 +00:00
|
|
|
const FORMAT = 1 << 0;
|
|
|
|
const CLIPPY = 1 << 1;
|
|
|
|
const COMPILE_FAIL = 1 << 2;
|
|
|
|
const TEST = 1 << 3;
|
|
|
|
const DOC_TEST = 1 << 4;
|
|
|
|
const DOC_CHECK = 1 << 5;
|
|
|
|
const BENCH_CHECK = 1 << 6;
|
|
|
|
const EXAMPLE_CHECK = 1 << 7;
|
|
|
|
const COMPILE_CHECK = 1 << 8;
|
|
|
|
const CFG_CHECK = 1 << 9;
|
|
|
|
const TEST_CHECK = 1 << 10;
|
2022-05-02 19:13:34 +00:00
|
|
|
}
|
|
|
|
|
2024-04-10 14:29:05 +00:00
|
|
|
/// A collection of flags that can modify how checks are run.
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
|
|
|
|
struct Flag: u32 {
|
2024-04-10 14:29:05 +00:00
|
|
|
/// Forces certain checks to continue running even if they hit an error.
|
|
|
|
const KEEP_GOING = 1 << 0;
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// None of the CI tests require any information at runtime other than the options that have been set,
|
|
|
|
// which is why all of these are 'static; we could easily update this to use more flexible types.
|
|
|
|
struct CITest<'a> {
|
|
|
|
/// The command to execute
|
|
|
|
command: Cmd<'a>,
|
|
|
|
|
|
|
|
/// The message to display if the test command fails
|
|
|
|
failure_message: &'static str,
|
|
|
|
|
|
|
|
/// The subdirectory path to run the test command within
|
|
|
|
subdir: Option<&'static str>,
|
|
|
|
|
|
|
|
/// Environment variables that need to be set before the test runs
|
|
|
|
env_vars: Vec<(&'static str, &'static str)>,
|
|
|
|
}
|
|
|
|
|
2021-02-22 08:42:19 +00:00
|
|
|
fn main() {
|
2021-03-11 00:27:30 +00:00
|
|
|
// When run locally, results may differ from actual CI runs triggered by
|
|
|
|
// .github/workflows/ci.yml
|
2021-02-22 08:42:19 +00:00
|
|
|
// - Official CI runs latest stable
|
|
|
|
// - Local runs use whatever the default Rust is locally
|
|
|
|
|
Warn when passing invalid argument to CI (#5858)
Example:
```sh
cargo run -p ci -- lint
Invalid argument: "lint".
Enter one of: lints, test, doc, compile, format, clippy, compile-fail, bench-check, example-check, doc-check, doc-test.
```
Co-authored-by: devil-ira <justthecooldude@gmail.com>
2022-09-02 12:18:45 +00:00
|
|
|
let arguments = [
|
|
|
|
("lints", Check::FORMAT | Check::CLIPPY),
|
|
|
|
("test", Check::TEST),
|
|
|
|
("doc", Check::DOC_TEST | Check::DOC_CHECK),
|
|
|
|
(
|
|
|
|
"compile",
|
2024-04-01 19:59:08 +00:00
|
|
|
Check::COMPILE_FAIL
|
|
|
|
| Check::BENCH_CHECK
|
|
|
|
| Check::EXAMPLE_CHECK
|
|
|
|
| Check::COMPILE_CHECK
|
|
|
|
| Check::TEST_CHECK,
|
Warn when passing invalid argument to CI (#5858)
Example:
```sh
cargo run -p ci -- lint
Invalid argument: "lint".
Enter one of: lints, test, doc, compile, format, clippy, compile-fail, bench-check, example-check, doc-check, doc-test.
```
Co-authored-by: devil-ira <justthecooldude@gmail.com>
2022-09-02 12:18:45 +00:00
|
|
|
),
|
|
|
|
("format", Check::FORMAT),
|
|
|
|
("clippy", Check::CLIPPY),
|
|
|
|
("compile-fail", Check::COMPILE_FAIL),
|
|
|
|
("bench-check", Check::BENCH_CHECK),
|
|
|
|
("example-check", Check::EXAMPLE_CHECK),
|
2024-04-01 19:59:08 +00:00
|
|
|
("test-check", Check::TEST_CHECK),
|
2024-02-25 15:19:27 +00:00
|
|
|
("cfg-check", Check::CFG_CHECK),
|
Warn when passing invalid argument to CI (#5858)
Example:
```sh
cargo run -p ci -- lint
Invalid argument: "lint".
Enter one of: lints, test, doc, compile, format, clippy, compile-fail, bench-check, example-check, doc-check, doc-test.
```
Co-authored-by: devil-ira <justthecooldude@gmail.com>
2022-09-02 12:18:45 +00:00
|
|
|
("doc-check", Check::DOC_CHECK),
|
|
|
|
("doc-test", Check::DOC_TEST),
|
|
|
|
];
|
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let flag_arguments = [("--keep-going", Flag::KEEP_GOING)];
|
|
|
|
|
|
|
|
// Parameters are parsed disregarding their order. Note that the first arg is generally the name of
|
|
|
|
// the executable, so it is ignored. Any parameter may either be a flag or the name of a battery of tests
|
|
|
|
// to include.
|
|
|
|
let (mut checks, mut flags) = (Check::empty(), Flag::empty());
|
2024-04-10 14:29:05 +00:00
|
|
|
|
|
|
|
// Skip first argument, which is usually the path of the executable.
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
for arg in std::env::args().skip(1) {
|
|
|
|
if let Some((_, check)) = arguments.iter().find(|(check_arg, _)| *check_arg == arg) {
|
|
|
|
// Note that this actually adds all of the constituent checks to the test suite.
|
|
|
|
checks.insert(*check);
|
2024-04-10 14:29:05 +00:00
|
|
|
} else if let Some((_, flag)) = flag_arguments.iter().find(|(flag_arg, _)| *flag_arg == arg)
|
|
|
|
{
|
|
|
|
flags.insert(*flag);
|
|
|
|
} else {
|
|
|
|
// We encountered an invalid parameter:
|
|
|
|
eprintln!(
|
|
|
|
"Invalid argument: {arg:?}.\n\
|
|
|
|
Valid parameters: {}.",
|
|
|
|
// Skip first argument so it can be added in fold(), when displaying as a comma separated list.
|
|
|
|
arguments[1..]
|
|
|
|
.iter()
|
|
|
|
.map(|(s, _)| s)
|
|
|
|
.chain(flag_arguments.iter().map(|(s, _)| s))
|
|
|
|
.fold(arguments[0].0.to_owned(), |c, v| c + ", " + v)
|
|
|
|
);
|
|
|
|
|
|
|
|
return;
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// If no checks are specified, we run every check
|
|
|
|
if checks.is_empty() {
|
|
|
|
checks = Check::all();
|
|
|
|
}
|
2021-02-22 08:42:19 +00:00
|
|
|
|
2022-05-30 16:21:03 +00:00
|
|
|
let sh = Shell::new().unwrap();
|
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
// Each check contains a 'battery' (vector) that can include more than one command, but almost all of them
|
|
|
|
// just contain a single command.
|
|
|
|
let mut test_suite: BTreeMap<Check, Vec<CITest>> = BTreeMap::new();
|
|
|
|
|
|
|
|
if checks.contains(Check::FORMAT) {
|
2022-05-02 19:13:34 +00:00
|
|
|
// See if any code needs to be formatted
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
test_suite.insert(
|
|
|
|
Check::FORMAT,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo fmt --all -- --check"),
|
|
|
|
failure_message: "Please run 'cargo fmt --all' to format your code.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![],
|
|
|
|
}],
|
|
|
|
);
|
2022-05-02 19:13:34 +00:00
|
|
|
}
|
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::CLIPPY) {
|
2022-05-02 19:13:34 +00:00
|
|
|
// See if clippy has any complaints.
|
|
|
|
// - Type complexity must be ignored because we use huge templates for queries
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
test_suite.insert(
|
|
|
|
Check::CLIPPY,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(
|
|
|
|
sh,
|
|
|
|
"cargo clippy --workspace --all-targets --all-features -- -Dwarnings"
|
|
|
|
),
|
|
|
|
failure_message: "Please fix clippy errors in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![],
|
|
|
|
}],
|
|
|
|
);
|
|
|
|
}
|
|
|
|
|
|
|
|
if checks.contains(Check::COMPILE_FAIL) {
|
|
|
|
let mut args = vec!["--target-dir", "../../target"];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--no-fail-fast");
|
bevy_derive: Add `#[deref]` attribute (#8552)
# Objective
Bevy code tends to make heavy use of the [newtype](
https://doc.rust-lang.org/rust-by-example/generics/new_types.html)
pattern, which is why we have a dedicated derive for
[`Deref`](https://doc.rust-lang.org/std/ops/trait.Deref.html) and
[`DerefMut`](https://doc.rust-lang.org/std/ops/trait.DerefMut.html).
This derive works for any struct with a single field:
```rust
#[derive(Component, Deref, DerefMut)]
struct MyNewtype(usize);
```
One reason for the single-field limitation is to prevent confusion and
footguns related that would arise from allowing multi-field structs:
<table align="center">
<tr>
<th colspan="2">
Similar structs, different derefs
</th>
</tr>
<tr>
<td>
```rust
#[derive(Deref, DerefMut)]
struct MyStruct {
foo: usize, // <- Derefs usize
bar: String,
}
```
</td>
<td>
```rust
#[derive(Deref, DerefMut)]
struct MyStruct {
bar: String, // <- Derefs String
foo: usize,
}
```
</td>
</tr>
<tr>
<th colspan="2">
Why `.1`?
</th>
</tr>
<tr>
<td colspan="2">
```rust
#[derive(Deref, DerefMut)]
struct MyStruct(Vec<usize>, Vec<f32>);
let mut foo = MyStruct(vec![123], vec![1.23]);
// Why can we skip the `.0` here?
foo.push(456);
// But not here?
foo.1.push(4.56);
```
</td>
</tr>
</table>
However, there are certainly cases where it's useful to allow for
structs with multiple fields. Such as for structs with one "real" field
and one `PhantomData` to allow for generics:
```rust
#[derive(Deref, DerefMut)]
struct MyStruct<T>(
// We want use this field for the `Deref`/`DerefMut` impls
String,
// But we need this field so that we can make this struct generic
PhantomData<T>
);
// ERROR: Deref can only be derived for structs with a single field
// ERROR: DerefMut can only be derived for structs with a single field
```
Additionally, the possible confusion and footguns are mainly an issue
for newer Rust/Bevy users. Those familiar with `Deref` and `DerefMut`
understand what adding the derive really means and can anticipate its
behavior.
## Solution
Allow users to opt into multi-field `Deref`/`DerefMut` derives using a
`#[deref]` attribute:
```rust
#[derive(Deref, DerefMut)]
struct MyStruct<T>(
// Use this field for the `Deref`/`DerefMut` impls
#[deref] String,
// We can freely include any other field without a compile error
PhantomData<T>
);
```
This prevents the footgun pointed out in the first issue described in
the previous section, but it still leaves the possible confusion
surrounding `.0`-vs-`.#`. However, the idea is that by making this
behavior explicit with an attribute, users will be more aware of it and
can adapt appropriately.
---
## Changelog
- Added `#[deref]` attribute to `Deref` and `DerefMut` derives
2023-05-16 18:29:09 +00:00
|
|
|
}
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
|
|
|
|
// ECS Compile Fail Tests
|
|
|
|
// Run UI tests (they do not get executed with the workspace tests)
|
|
|
|
// - See crates/bevy_ecs_compile_fail_tests/README.md
|
|
|
|
|
|
|
|
// (These must be cloned because of move semantics in `cmd!`)
|
|
|
|
let args_clone = args.clone();
|
|
|
|
|
|
|
|
test_suite.insert(Check::COMPILE_FAIL, vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo test {args_clone...}"),
|
|
|
|
failure_message: "Compiler errors of the ECS compile fail tests seem to be different than expected! Check locally and compare rust versions.",
|
|
|
|
subdir: Some("crates/bevy_ecs_compile_fail_tests"),
|
|
|
|
env_vars: vec![],
|
|
|
|
}]);
|
|
|
|
|
|
|
|
// Reflect Compile Fail Tests
|
|
|
|
// Run tests (they do not get executed with the workspace tests)
|
|
|
|
// - See crates/bevy_reflect_compile_fail_tests/README.md
|
|
|
|
let args_clone = args.clone();
|
|
|
|
|
|
|
|
test_suite.entry(Check::COMPILE_FAIL).and_modify(|tests| tests.push( CITest {
|
|
|
|
command: cmd!(sh, "cargo test {args_clone...}"),
|
|
|
|
failure_message: "Compiler errors of the Reflect compile fail tests seem to be different than expected! Check locally and compare rust versions.",
|
|
|
|
subdir: Some("crates/bevy_reflect_compile_fail_tests"),
|
|
|
|
env_vars: vec![],
|
|
|
|
}));
|
|
|
|
|
|
|
|
// Macro Compile Fail Tests
|
|
|
|
// Run tests (they do not get executed with the workspace tests)
|
|
|
|
// - See crates/bevy_macros_compile_fail_tests/README.md
|
|
|
|
let args_clone = args.clone();
|
|
|
|
|
|
|
|
test_suite.entry(Check::COMPILE_FAIL).and_modify(|tests| tests.push( CITest {
|
|
|
|
command: cmd!(sh, "cargo test {args_clone...}"),
|
|
|
|
failure_message: "Compiler errors of the macros compile fail tests seem to be different than expected! Check locally and compare rust versions.",
|
|
|
|
subdir: Some("crates/bevy_macros_compile_fail_tests"),
|
|
|
|
env_vars: vec![],
|
|
|
|
}));
|
2021-11-13 22:43:19 +00:00
|
|
|
}
|
2022-02-03 04:25:45 +00:00
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::TEST) {
|
2022-05-02 19:13:34 +00:00
|
|
|
// Run tests (except doc tests and without building examples)
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let mut args = vec!["--workspace", "--lib", "--bins", "--tests", "--benches"];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--no-fail-fast");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::TEST,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo test {args...}"),
|
|
|
|
failure_message: "Please fix failing tests in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![],
|
|
|
|
}],
|
|
|
|
);
|
2022-05-02 19:13:34 +00:00
|
|
|
}
|
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::DOC_TEST) {
|
2022-05-02 19:13:34 +00:00
|
|
|
// Run doc tests
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let mut args = vec!["--workspace", "--doc"];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--no-fail-fast");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::DOC_TEST,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo test {args...}"),
|
|
|
|
failure_message: "Please fix failing doc-tests in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![],
|
|
|
|
}],
|
|
|
|
);
|
2022-05-02 19:13:34 +00:00
|
|
|
}
|
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::DOC_CHECK) {
|
2022-05-02 19:13:34 +00:00
|
|
|
// Check that building docs work and does not emit warnings
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let mut args = vec![
|
|
|
|
"--workspace",
|
|
|
|
"--all-features",
|
|
|
|
"--no-deps",
|
|
|
|
"--document-private-items",
|
|
|
|
];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--keep-going");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::DOC_CHECK,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo doc {args...}"),
|
|
|
|
failure_message: "Please fix doc warnings in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![("RUSTDOCFLAGS", "-D warnings")],
|
|
|
|
}],
|
|
|
|
);
|
2022-05-02 19:13:34 +00:00
|
|
|
}
|
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::BENCH_CHECK) {
|
2022-05-02 19:13:34 +00:00
|
|
|
// Check that benches are building
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let mut args = vec!["--benches", "--target-dir", "../target"];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--keep-going");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::BENCH_CHECK,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo check {args...}"),
|
|
|
|
failure_message: "Failed to check the benches.",
|
|
|
|
subdir: Some("benches"),
|
|
|
|
env_vars: vec![],
|
|
|
|
}],
|
|
|
|
);
|
2022-05-02 19:13:34 +00:00
|
|
|
}
|
2022-02-03 04:25:45 +00:00
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::EXAMPLE_CHECK) {
|
2022-05-02 19:13:34 +00:00
|
|
|
// Build examples and check they compile
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let mut args = vec!["--workspace", "--examples"];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--keep-going");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::EXAMPLE_CHECK,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo check {args...}"),
|
|
|
|
failure_message: "Please fix compiler errors for examples in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![],
|
|
|
|
}],
|
|
|
|
);
|
2022-02-03 04:25:45 +00:00
|
|
|
}
|
2022-05-06 19:29:44 +00:00
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::COMPILE_CHECK) {
|
2023-01-25 03:20:58 +00:00
|
|
|
// Build bevy and check that it compiles
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let mut args = vec!["--workspace"];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--keep-going");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::COMPILE_CHECK,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo check {args...}"),
|
|
|
|
failure_message: "Please fix compiler errors in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![],
|
|
|
|
}],
|
|
|
|
);
|
2022-05-06 19:29:44 +00:00
|
|
|
}
|
2024-02-25 15:19:27 +00:00
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
if checks.contains(Check::CFG_CHECK) {
|
2024-02-25 15:19:27 +00:00
|
|
|
// Check cfg and imports
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
let mut args = vec!["-Zcheck-cfg", "--workspace"];
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--keep-going");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::CFG_CHECK,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo +nightly check {args...}"),
|
|
|
|
failure_message: "Please fix failing cfg checks in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: vec![("RUSTFLAGS", "-D warnings")],
|
|
|
|
}],
|
|
|
|
);
|
|
|
|
}
|
|
|
|
|
2024-04-01 19:59:08 +00:00
|
|
|
if checks.contains(Check::TEST_CHECK) {
|
|
|
|
let mut args = vec!["--workspace", "--tests"];
|
|
|
|
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
args.push("--keep-going");
|
|
|
|
}
|
|
|
|
|
|
|
|
test_suite.insert(
|
|
|
|
Check::TEST_CHECK,
|
|
|
|
vec![CITest {
|
|
|
|
command: cmd!(sh, "cargo check {args...}"),
|
|
|
|
failure_message: "Please fix compiler examples for tests in output above.",
|
|
|
|
subdir: None,
|
|
|
|
env_vars: Vec::new(),
|
|
|
|
}],
|
|
|
|
);
|
|
|
|
}
|
|
|
|
|
Quality-of-life updates for running CI locally (#12242)
# Objective
- Make running CI locally relatively less painful by allowing
continuation after failure
- Fix #12206
## Solution
Previously, `ci` would halt after encounting a failure (or shortly
thereafter). For instance, if you ran `cargo run -p ci` and a single
test within a CI check failed somewhere in the middle, you would never
even reach many of the other tests within the check and certainly none
of the CI checks that came afterward. This was annoying; if I am fixing
issues with CI tests, I want to just see all of the issues up-front
instead of having to rerun CI every time I fix something to see the next
error.
Furthermore, it is not infrequent (because of subtle
configuration/system differences) to encounter spurious CI failures
locally; previously, when these occurred, they would make the process of
running CI largely pointless, since you would have to push your branch
in order to see your actual CI failures from the automated testing
service.
Now, when running `ci` locally, we have a couple new tools to ameliorate
these and provide a smoother experience:
- Firstly, there is a new optional flag `--keep-going` that can be
passed while calling `ci` (e.g. `cargo run -p ci -- doc --keep-going`).
It has the following effects:
- It causes the `--keep-going` flag to be passed to the script's `cargo
doc` and `cargo check` invocations, so that they do not stop when they
encounter an error in a single module; instead, they keep going (!) and
find errors subsequently.
- It causes the `--no-fail-fast` flag to be passed to the script's
`cargo test` invocations, to similar effect.
- Finally, it causes `ci` itself not to abort after a single check
fails; instead, it will run every check that was invoked.
Thus, for instance, `cargo run -p ci -- --keep-going` will run every CI
check even if it encounters intermediate failures, and every such check
will itself be run to completion.
- Secondly, we now allow multiple ordinary arguments to be passed to
`ci`. For instance, `cargo -p ci -- doc test` now executes both the
'doc' checks and the 'test' checks. This allows the caller to run the
tests they care about with fewer invocations of `ci`.
As of this PR, automated testing will remain unchanged.
---
## Changelog
- tools/ci/src/main.rs refactored into staging and execution steps,
since the previous control flow did not naturally support continuing
after failure.
- Added "--keep-going" flag to `ci`.
- Added support for invoking `ci` with multiple arguments.
---
## Discussion
### Design considerations
I had originally split this into separate flags that controlled:
1. whether `--keep-going`/`--no-fail-fast` would be passed to the
constituent commands
2. whether `ci` would continue after a component test failed
However, I decided to merge these two behaviors, since I think that, if
you're in the business of seeing all the errors, you'll probably want to
actually see all of the errors. One slightly annoying thing, however,
about the new behavior with `--keep-going`, is that you will sometimes
find yourself wishing that the script would pause or something, since it
tends to fill the screen with a lot of junk. I have found that sending
stdout to /dev/null helps quite a bit, but I don't think `cargo fmt` or
`cargo clippy` actually write to stderr, so you need to be cognizant of
that (and perhaps invoke the lints separately).
~~Next, I'll mention that I feel somewhat strongly that the current
behavior of `ci` for automated testing should remain the same, since its
job is more like detecting that errors exist rather than finding all of
them.~~ (I was convinced this could have value.) Orthogonally, there is
the question of whether or not we might want to make this (or something
similar) actually the default behavior and make the automated test
script invoke some optional flags — it doesn't have to type with its
hands, after all. I'm not against that, but I don't really want to rock
the boat much more with this PR, since anyone who looks at the diff
might already be a little incensed.
2024-03-04 22:03:02 +00:00
|
|
|
// Actually run the tests:
|
|
|
|
|
|
|
|
let mut failed_checks: Check = Check::empty();
|
|
|
|
let mut failure_message: String = String::new();
|
|
|
|
|
|
|
|
// In KEEP_GOING-mode, we save all errors until the end; otherwise, we just
|
|
|
|
// panic with the given message for test failure.
|
|
|
|
fn fail(
|
|
|
|
current_check: Check,
|
|
|
|
failure_message: &'static str,
|
|
|
|
failed_checks: &mut Check,
|
|
|
|
existing_fail_message: &mut String,
|
|
|
|
flags: &Flag,
|
|
|
|
) {
|
|
|
|
if flags.contains(Flag::KEEP_GOING) {
|
|
|
|
failed_checks.insert(current_check);
|
|
|
|
if !existing_fail_message.is_empty() {
|
|
|
|
existing_fail_message.push('\n');
|
|
|
|
}
|
|
|
|
existing_fail_message.push_str(failure_message);
|
|
|
|
} else {
|
|
|
|
panic!("{failure_message}");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
for (check, battery) in test_suite {
|
|
|
|
for ci_test in battery {
|
|
|
|
// If the CI test is to be executed in a subdirectory, we move there before running the command
|
|
|
|
let _subdir_hook = ci_test.subdir.map(|path| sh.push_dir(path));
|
|
|
|
|
|
|
|
// Actually run the test, setting environment variables temporarily
|
|
|
|
if ci_test.command.envs(ci_test.env_vars).run().is_err() {
|
|
|
|
fail(
|
|
|
|
check,
|
|
|
|
ci_test.failure_message,
|
|
|
|
&mut failed_checks,
|
|
|
|
&mut failure_message,
|
|
|
|
&flags,
|
|
|
|
);
|
|
|
|
}
|
|
|
|
// ^ This must run while `_subdir_hook` is in scope; it is dropped on loop iteration.
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if !failed_checks.is_empty() {
|
|
|
|
panic!(
|
|
|
|
"One or more CI checks failed.\n\
|
|
|
|
{failure_message}"
|
|
|
|
);
|
2024-02-25 15:19:27 +00:00
|
|
|
}
|
2021-02-22 08:42:19 +00:00
|
|
|
}
|