Anyone has contributed code to the project can become a member of the project and have the write permission to issues and doc repositories.
At the early stage of this project, we will use a simple model to promote members to maintainers. Maintainers will have all the permissions of this project, only the first 2 maintainers are granted by the owner, the standard is whether the member is good enough to review others' code, then we will start to elect new maintainers by voting in the public issue. If no one votes down and 2/3 votes up then an election passes.
Check here.
When we talk about type in the doc we use gopls symbol query syntax. For example, when we say rod.Page.PDF
, you can run the below to locate the file and line of it:
gopls workspace_symbol -matcher=fuzzy rod.Page.PDF$
cdp
: It's short for Chrome Devtools Protocol
Here's the common start process of rod:
-
Try to connect to a Devtools endpoint (WebSocket), if not found try to launch a local browser, if still not found try to download one, then connect again. The lib to handle it is launcher.
-
Use the JSON-RPC to talk to the Devtools endpoint to control the browser. The lib handles it is cdp.
-
Use the type definitions of the JSON-RPC to perform high-level actions. The lib handles it is proto.
Object model:
First, launch a test shell for rod:
go run ./lib/utils/shell
Then, no magic, just go test
. Read the test template rod_test.go to get started.
The entry point of tests is setup_test.go. All the test helpers are defined in it.
The cdp
requests of each test will be recorded and output to folder tmp/cdp-log
, the CI will store them as
artifacts so that we can download
them for debugging.
Usually, you only need to run the tests that you are working on, for example:
go test -run=^TestClick$
The above will only run TestClick.
go test -rod=show
Check defaults for other available options.
You can run all commands inside Docker so that you don't have to install all the development dependencies. Check Use Docker for development for more info.
go generate # only required for first time
go run ./lib/utils/lint
If the code coverage is less than 100%, the CI will fail.
Learn the basics first.
To visually see the coverage report you can run something like this:
go test -coverprofile=coverage.out ./...
go tool cover -html=coverage.out
It will open a web page to tell you which line is not covered.
To cover the error branch of the code we usually intercept cdp calls. There are several helper functions for it:
rod_test.MockClient.stubCounter
rod_test.MockClient.stub
rod_test.MockClient.stubErr
-
Build the test image:
docker build -t rod -f lib/docker/dev.Dockerfile .
-
Run a container with and mount the cache volume to it:
docker run -v $(pwd):/t --name rod -it rod bash
-
Open another terminal, copy your global git-ignore file to the container:
docker cp ~/.gitignore_global rod:/root/
-
Run lint in the container:
go run ./lib/utils/lint
-
Run tests in the container:
go test
-
After you exit the container with
exit
, you can restart it by:docker start -i rod
We use .github/workflows/docker.yml
to automate it.
Because parallel execution will pollution the global goroutine stack, by default, the goroutine leak detection for each test will be disabled, but the detection for the whole test program will still work as well. To enable detection for each test, just use go test -parallel=1
.
Run go mod vendor
to create a local mirror of dependencies.
The Golang compiler will use the libs under vendor
folder as a priority.
For example, we can modify file ./vendor/github.com/ysmood/goob/goob.go
to debug, such as add some extra logs.
All conversations in Github issues, PRs, etc. should be summarized into code comments so that this project is not deep coupled with Github service.
The commit message follows the rules here. We don't use rules like Conventional Commits because it's hard for beginners to write correct commit messages. It will encourage reviewers to spend more time on high-level problems, not the details. We also want to reduce the overhead when reading the git-blame, for example, fix: correct minor typos in code
is the same as fix minor typos in code
, there's no need to repeat content in the title line.