-
-
Save LukeMathWalker/d98fa8d0fc5394b347adf734ef0e85ec to your computer and use it in GitHub Desktop.
image: "rust:latest" | |
default: | |
before_script: | |
- rustc --version | |
- cargo --version | |
stages: | |
- test | |
test-code: | |
stage: test | |
script: | |
- cargo test | |
- cargo install cargo-tarpaulin | |
- cargo tarpaulin --ignore-tests | |
lint-code: | |
stage: test | |
script: | |
- rustup component add clippy | |
- cargo clippy -- -D warnings | |
format-code: | |
stage: test | |
script: | |
- rustup component add rustfmt | |
- cargo fmt -- --check | |
audit-code: | |
stage: test | |
script: | |
- cargo install cargo-audit | |
- cargo audit |
Thanks - updated @devonh!
What else needs to be added to the Gitlab version to account for SQLX's compile time checks?
To consider coverage in Merge Requests, the test-code
job should be:
test-code:
stage: test
artifacts:
when: always
reports:
cobertura: "cobertura.xml"
coverage: '/\d+\.\d+% coverage, /'
script:
- cargo test
- cargo install cargo-tarpaulin
- cargo tarpaulin --ignore-tests --out Xml
image
can be moved under the default
part, to allow individual jobs to use different images (for example, if you want to compile in different Rust versions), leaving it like this:
default:
image: rust:latest
before_script:
- rustc --version
- cargo --version
To add caching, you can use the following on top of the file:
cache:
paths:
- target/
I don't know if Rust requires to cache any other directory, but you can easily add any other directory under the paths
piece (or exclude directories if it may be a problem)
Ran into:
$ cargo test
Compiling zero2prod v0.1.0 (/builds/SniffleEU/zero2prod)
error: linkerclang
not found = note: No such file or directory (os error 2)
error: could not compilezero2prod
due to previous error
Cleaning up project directory and file based variables 00:01
ERROR: Job failed: exit code 1
In the test-code block.
As I've never tried gitlab CI before my workaround is crude, but changing from:
test-code:
stage: test
script:
- cargo test
- cargo install cargo-tarpaulin
- cargo tarpaulin --ignore-tests
to
test-code:
stage: test
script:
- apt update
- apt install lld clang -y
- cargo test
- cargo install cargo-tarpaulin
- cargo tarpaulin --ignore-tests
Fixes the issue and runs the block
@Arevjensen Gitlab CI runs in a docker container, in this case, rust:latest
as specified at the top of the file here. As you've discovered the rust container doesn't include lld or clang by default.
When doing it this way, you're installing these each and every time. I've gone ahead and bundled all the dependencies as of Chapter 1 into a single docker image.
If you'd like you can replace image: "rust-latest"
with image: "aestusvitae/rust-ci:latest"
and remove the installs (apt
, cargo install
, rustup component add
) from your CI scripts, and CI runs should be noticeably quicker.
You can even build the image yourself if you're so inclined. (or if I fail to keep mine up to date) The Dockerfile is very simple.
What else needs to be added to the Gitlab version to account for SQLX's compile time checks?
@Valmirius , you can use Gitlab CI services for the tests and coverage jobs:
tests:
stage: test
services:
- postgres:14
variables:
POSTGRES_DB: newsletter
POSTGRES_USER: postgres
POSTGRES_PASSWORD: password
POSTGRES_HOST: postgres
POSTGRES_PORT: 5432
DATABASE_URL: postgres://postgres:password@postgres:5432/newsletter
script:
- SKIP_DOCKER=true ./scripts/init_db.sh
- cargo test
coverage:
stage: test
services:
- postgres:14
variables:
POSTGRES_DB: newsletter
POSTGRES_USER: postgres
POSTGRES_PASSWORD: password
POSTGRES_HOST: postgres
POSTGRES_PORT: 5432
DATABASE_URL: postgres://postgres:password@postgres:5432/newsletter
script:
- SKIP_DOCKER=true ./scripts/init_db.sh
- cargo tarpaulin --ignore-tests
the way it works is that gitlab runner instantiates another container (the service), and you can make requests to it with a hostname the same as the service name. The variables
section allows you to set whatever the service requires to work. At hub.docker.com for postgres image, you can find that you require those POSTGRES_
suffixed variables.
I added DATABASE_URL
variable to override what the .env
file has as default, you may recall that it uses localhost as hostname. For that t o work I needed to add the following line to tests/health_check.rs so that the configuration read from the yaml overrides the url string that gets built when configuration is read using yaml:
let connection_string = configuration.database.connection_string();
+ let env_connection_string = std::env::var("DATABASE_URL").unwrap_or(connection_string);
let mut connection = PgConnection::connect(&env_connection_string)
I know that config allows env variables to get this same effect, but haven't found how to make it work. This solution at least works for me currently.
Hey! To parallelize things you could update this to something like the following: