nushell/.github/workflows/ci.yml

208 lines
6.3 KiB
YAML
Raw Normal View History

on:
pull_request:
push: # Run CI on the main branch after every merge. This is important to fill the GitHub Actions cache in a way that pull requests can see it
branches:
- main
2021-09-04 07:52:28 +00:00
2022-04-24 13:03:21 +00:00
name: continuous-integration
2021-09-04 07:52:28 +00:00
jobs:
nu-fmt-clippy:
2021-09-04 07:52:28 +00:00
strategy:
fail-fast: true
2021-09-04 07:52:28 +00:00
matrix:
# Pinning to Ubuntu 20.04 because building on newer Ubuntu versions causes linux-gnu
# builds to link against a too-new-for-many-Linux-installs glibc version. Consider
# revisiting this when 20.04 is closer to EOL (April 2025)
platform: [windows-latest, macos-latest, ubuntu-20.04]
style: [default, dataframe]
2021-09-04 07:52:28 +00:00
rust:
- stable
include:
- style: default
flags: ""
- style: dataframe
flags: "--features=dataframe "
exclude:
# only test dataframes on Ubuntu (the fastest platform)
- platform: windows-latest
style: dataframe
- platform: macos-latest
style: dataframe
2021-09-04 07:52:28 +00:00
2021-10-11 19:19:43 +00:00
runs-on: ${{ matrix.platform }}
env:
NUSHELL_CARGO_TARGET: ci
2021-10-11 19:19:43 +00:00
2021-09-04 07:52:28 +00:00
steps:
- uses: actions/checkout@v3
2021-09-04 07:52:28 +00:00
2022-10-30 07:24:10 +00:00
- name: Setup Rust toolchain and cache
uses: actions-rust-lang/setup-rust-toolchain@v1.4.4
- name: cargo fmt
run: cargo fmt --all -- --check
- name: Clippy
run: cargo clippy --workspace ${{ matrix.flags }}--exclude nu_plugin_* -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect
2021-09-04 08:05:36 +00:00
nu-tests:
env:
NUSHELL_CARGO_TARGET: ci
strategy:
fail-fast: true
matrix:
platform: [windows-latest, macos-latest, ubuntu-20.04]
style: [default, dataframe]
rust:
- stable
include:
- style: default
flags: ""
- style: dataframe
flags: "--features=dataframe"
exclude:
# only test dataframes on Ubuntu (the fastest platform)
- platform: windows-latest
style: dataframe
- platform: macos-latest
style: dataframe
runs-on: ${{ matrix.platform }}
steps:
- uses: actions/checkout@v3
2022-10-30 07:24:10 +00:00
- name: Setup Rust toolchain and cache
uses: actions-rust-lang/setup-rust-toolchain@v1.4.4
- name: Tests
run: cargo test --workspace --profile ci --exclude nu_plugin_* ${{ matrix.flags }}
standard library: bring the tests into the main CI (#8525) Should close one of the tasks in #8450. # Description > **Note** > in order of appearance in the global diff - 1b7497c41966306aa3103a95a9b5ef5df7111ee4 adds the `std-tests` job to the CI which 1. installs `nushell` in the runner 2. run the `tests.nu` module > see `open .github/workflows/ci.yml | get jobs.std-tests | to yaml` - [`ec85b6fd`..`9c122115`](ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629..9c12211564ca8ee90ed65ae45776dccb8f8e4ef1) is where all the magic happens => see below - :test_tube: 799c7eb7fd5f140289b36b9dbc00329c50e2fbda introduces some bugs and failing test to see how the CI behaves => see how the [tests failed](https://github.com/nushell/nushell/actions/runs/4460098237/jobs/7833018256) as expected :x: - :test_tube: and c3de1fafb5c5313e30c08c9ca57e09df33b61b74 reverts the failing tests, i.e. the previous commit, leaving a standard library whose tests all pass :tada: => see the [tests passing](https://github.com/nushell/nushell/actions/runs/4460153434/jobs/7833110719?pr=8525#step:5:1) now :heavy_check_mark: ## the changes to the runner > see [`ec85b6fd`..`9c122115`](ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629..9c12211564ca8ee90ed65ae45776dccb8f8e4ef1) the issue with the previous runner was the following: the clever trick of using `nu -c "use ...; test"` did print the errors when occuring but they did not capture the true "failure", i.e. in all cases the `$env.LAST_EXIT_CODE` was set to `0`, never stopping the CI when a test failed :thinking: i first tried to `try` / `catch` the error in ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629 which kinda worked but only throw a single error, the first one i thought it was not the best and started thinking about a solution to have a complete report of all failing tests, at once, to avoid running the CI multiple times! the easiest solution i found was the one i implemented in 9c12211564ca8ee90ed65ae45776dccb8f8e4ef1 > **Warning** > this changes the structure of the runner quite a bit, but the `for` loops where annoying to manipulate structured data and allow the runner to draw a complete report... now the runner does the following - compute the list of all available tests in a table with the `file`, `module` and `name` columns (first part of the pipe until `flatten` and `rename`) - run the tests one by one computing the new `pass` column - with a `log info` - captures the failing ones => puts `true` in `pass` if the test passes, `false` otherwise - if at least one test has failed, throw a single error with the list of failing tests ### hope you'll like it :relieved: # User-Facing Changes ``` $nothing ``` # Tests + Formatting the standard tests now return a true error that will stop the CI # After Submitting ``` $nothing ```
2023-03-25 18:29:08 +00:00
std-lib-and-python-virtualenv:
env:
standard library: bring the tests into the main CI (#8525) Should close one of the tasks in #8450. # Description > **Note** > in order of appearance in the global diff - 1b7497c41966306aa3103a95a9b5ef5df7111ee4 adds the `std-tests` job to the CI which 1. installs `nushell` in the runner 2. run the `tests.nu` module > see `open .github/workflows/ci.yml | get jobs.std-tests | to yaml` - [`ec85b6fd`..`9c122115`](ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629..9c12211564ca8ee90ed65ae45776dccb8f8e4ef1) is where all the magic happens => see below - :test_tube: 799c7eb7fd5f140289b36b9dbc00329c50e2fbda introduces some bugs and failing test to see how the CI behaves => see how the [tests failed](https://github.com/nushell/nushell/actions/runs/4460098237/jobs/7833018256) as expected :x: - :test_tube: and c3de1fafb5c5313e30c08c9ca57e09df33b61b74 reverts the failing tests, i.e. the previous commit, leaving a standard library whose tests all pass :tada: => see the [tests passing](https://github.com/nushell/nushell/actions/runs/4460153434/jobs/7833110719?pr=8525#step:5:1) now :heavy_check_mark: ## the changes to the runner > see [`ec85b6fd`..`9c122115`](ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629..9c12211564ca8ee90ed65ae45776dccb8f8e4ef1) the issue with the previous runner was the following: the clever trick of using `nu -c "use ...; test"` did print the errors when occuring but they did not capture the true "failure", i.e. in all cases the `$env.LAST_EXIT_CODE` was set to `0`, never stopping the CI when a test failed :thinking: i first tried to `try` / `catch` the error in ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629 which kinda worked but only throw a single error, the first one i thought it was not the best and started thinking about a solution to have a complete report of all failing tests, at once, to avoid running the CI multiple times! the easiest solution i found was the one i implemented in 9c12211564ca8ee90ed65ae45776dccb8f8e4ef1 > **Warning** > this changes the structure of the runner quite a bit, but the `for` loops where annoying to manipulate structured data and allow the runner to draw a complete report... now the runner does the following - compute the list of all available tests in a table with the `file`, `module` and `name` columns (first part of the pipe until `flatten` and `rename`) - run the tests one by one computing the new `pass` column - with a `log info` - captures the failing ones => puts `true` in `pass` if the test passes, `false` otherwise - if at least one test has failed, throw a single error with the list of failing tests ### hope you'll like it :relieved: # User-Facing Changes ``` $nothing ``` # Tests + Formatting the standard tests now return a true error that will stop the CI # After Submitting ``` $nothing ```
2023-03-25 18:29:08 +00:00
NU_LOG_LEVEL: DEBUG
strategy:
fail-fast: true
matrix:
platform: [ubuntu-20.04, macos-latest, windows-latest]
rust:
- stable
2022-03-18 22:36:38 +00:00
py:
- py
runs-on: ${{ matrix.platform }}
2021-09-04 07:52:28 +00:00
steps:
- uses: actions/checkout@v3
2022-10-30 07:24:10 +00:00
- name: Setup Rust toolchain and cache
uses: actions-rust-lang/setup-rust-toolchain@v1.4.4
- name: Install Nushell
standard library: bring the tests into the main CI (#8525) Should close one of the tasks in #8450. # Description > **Note** > in order of appearance in the global diff - 1b7497c41966306aa3103a95a9b5ef5df7111ee4 adds the `std-tests` job to the CI which 1. installs `nushell` in the runner 2. run the `tests.nu` module > see `open .github/workflows/ci.yml | get jobs.std-tests | to yaml` - [`ec85b6fd`..`9c122115`](ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629..9c12211564ca8ee90ed65ae45776dccb8f8e4ef1) is where all the magic happens => see below - :test_tube: 799c7eb7fd5f140289b36b9dbc00329c50e2fbda introduces some bugs and failing test to see how the CI behaves => see how the [tests failed](https://github.com/nushell/nushell/actions/runs/4460098237/jobs/7833018256) as expected :x: - :test_tube: and c3de1fafb5c5313e30c08c9ca57e09df33b61b74 reverts the failing tests, i.e. the previous commit, leaving a standard library whose tests all pass :tada: => see the [tests passing](https://github.com/nushell/nushell/actions/runs/4460153434/jobs/7833110719?pr=8525#step:5:1) now :heavy_check_mark: ## the changes to the runner > see [`ec85b6fd`..`9c122115`](ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629..9c12211564ca8ee90ed65ae45776dccb8f8e4ef1) the issue with the previous runner was the following: the clever trick of using `nu -c "use ...; test"` did print the errors when occuring but they did not capture the true "failure", i.e. in all cases the `$env.LAST_EXIT_CODE` was set to `0`, never stopping the CI when a test failed :thinking: i first tried to `try` / `catch` the error in ec85b6fd3fc004cd94e3fada5c8e5fe2714fd629 which kinda worked but only throw a single error, the first one i thought it was not the best and started thinking about a solution to have a complete report of all failing tests, at once, to avoid running the CI multiple times! the easiest solution i found was the one i implemented in 9c12211564ca8ee90ed65ae45776dccb8f8e4ef1 > **Warning** > this changes the structure of the runner quite a bit, but the `for` loops where annoying to manipulate structured data and allow the runner to draw a complete report... now the runner does the following - compute the list of all available tests in a table with the `file`, `module` and `name` columns (first part of the pipe until `flatten` and `rename`) - run the tests one by one computing the new `pass` column - with a `log info` - captures the failing ones => puts `true` in `pass` if the test passes, `false` otherwise - if at least one test has failed, throw a single error with the list of failing tests ### hope you'll like it :relieved: # User-Facing Changes ``` $nothing ``` # Tests + Formatting the standard tests now return a true error that will stop the CI # After Submitting ``` $nothing ```
2023-03-25 18:29:08 +00:00
# prior to [*standard library: bring the tests into the main CI*](#8525)
# there was a `--profile ci` here in the `cargo install`, as well as
# `NUSHELL_CARGO_TARGET: ci` in the prelude above.
#
# this caused a "stackoverflow" error in the CI on windows,
# see [this failing job](https://github.com/nushell/nushell/actions/runs/4512034615/jobs/7944945590)
#
# the CI profile has been removed in 00b820de9021227d1910a9ea388297ee7aee308e
# as part of #8525.
run: cargo install --path . --locked --no-default-features
- name: Standard library tests
run: nu -c 'use std; std run-tests --path crates/nu-std'
2022-03-18 22:36:38 +00:00
- name: Setup Python
uses: actions/setup-python@v4
2022-03-18 22:36:38 +00:00
with:
python-version: "3.10"
2022-03-18 22:36:38 +00:00
- run: python -m pip install tox
# Get only the latest tagged version for stability reasons
2022-03-18 22:36:38 +00:00
- name: Install virtualenv
run: git clone https://github.com/pypa/virtualenv.git
2022-03-18 22:36:38 +00:00
shell: bash
2022-03-18 22:36:38 +00:00
- name: Test Nushell in virtualenv
run: |
cd virtualenv
# if we encounter problems with bleeding edge tests pin to the latest tag
# git checkout $(git describe --tags | cut -d - -f 1)
# We need to disable failing on coverage levels.
nu -c "open pyproject.toml | upsert tool.coverage.report.fail_under 1 | save patchproject.toml"
mv patchproject.toml pyproject.toml
tox -e ${{ matrix.py }} -- -k nushell
2022-03-18 22:36:38 +00:00
shell: bash
2022-05-02 07:20:57 +00:00
# Build+test plugins on their own, without the rest of Nu. This helps with CI parallelization and
# also helps test that the plugins build without any feature unification shenanigans
2022-05-02 07:20:57 +00:00
plugins:
env:
NUSHELL_CARGO_TARGET: ci
2022-05-02 07:20:57 +00:00
strategy:
fail-fast: true
2022-05-02 07:20:57 +00:00
matrix:
platform: [windows-latest, macos-latest, ubuntu-20.04]
2022-05-02 07:20:57 +00:00
rust:
- stable
runs-on: ${{ matrix.platform }}
steps:
- uses: actions/checkout@v3
2022-05-02 07:20:57 +00:00
2022-10-30 07:24:10 +00:00
- name: Setup Rust toolchain and cache
uses: actions-rust-lang/setup-rust-toolchain@v1.4.4
2022-05-02 07:20:57 +00:00
- name: Clippy
run: cargo clippy --package nu_plugin_* ${{ matrix.flags }} -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect
2022-05-02 07:20:57 +00:00
- name: Tests
run: cargo test --profile ci --package nu_plugin_*
nu-coverage:
env:
NUSHELL_CARGO_TARGET: ci
strategy:
fail-fast: true
matrix:
# disabled mac due to problems with merging coverage and similarity to linux
# disabled windows due to running out of disk space when having too many crates or tests
platform: [ubuntu-20.04] # windows-latest
rust:
- stable
runs-on: ${{ matrix.platform }}
steps:
- uses: actions/checkout@v3
- name: Setup Rust toolchain and cache
uses: actions-rust-lang/setup-rust-toolchain@v1.4.4
- name: Install cargo-llvm-cov
uses: taiki-e/install-action@cargo-llvm-cov
- name: Tests
shell: bash
run: |
source <(cargo llvm-cov show-env --export-prefix) # Set the environment variables needed to get coverage.
cargo llvm-cov clean --workspace # Remove artifacts that may affect the coverage results.
cargo build --workspace --profile ci
cargo test --workspace --profile ci
cargo llvm-cov report --profile ci --lcov --output-path lcov.info
- name: Upload coverage reports to Codecov with GitHub Action
uses: codecov/codecov-action@v3
with:
files: lcov.info