Skip to content

Latest commit

 

History

History
277 lines (247 loc) · 10.4 KB

Running and Writing Tests for Sia.md

File metadata and controls

277 lines (247 loc) · 10.4 KB

Running and Writing Tests for Sia

Improving test coverage is a great way to start contributing to Sia.

This guide focuses on how to write tests. To learn about making pull requests to submit the code you've written, see doc/Guide to Contributing to Sia.md. You should also read doc/Developers.md to learn about Sia code conventions and quality standards.

Table of Contents

Running tests for Sia

Go's comprehensive test package makes testing straightforward, particularly when you use the bundled tools included in the Sia makefile, including make test, make cover, make bench, and their variants.

Updating code before testing

If you just want to run existing tests on the codebase as is, you just need to pull the latest version of the original repo to your master branch. (If that sentence didn't make sense, go read doc/Guide to Contributing to Sia.md.)

# Make sure you are in the right directory.
$ cd $GOPATH/src/github.com/<your Github username>/Sia
# Also make sure you're working with the right branch.
$ git checkout master
# Pull latest changes from origin, the original Sia repo. 
$ git pull origin master
# Update your fork of the repo, which should be set up as a remote.
$ git push <remote>  master

If you want to run tests on the new code you've added, first make sure the rest of the code is up to date. New code should be on its own branch (again, see doc/Guide to Contributing to Sia.md).

# Make sure you are in the right directory.
$ cd $GOPATH/src/github.com/<your Github username>/Sia
# Checkout the branch you made the changes on.
$ git checkout <branch name>
# Stash any tracked but uncommitted changes.
$ git stash
# Then switch back to `master` and update it to match the original repo.
$ git checkout master
$ git pull origin master
# Update your fork of the repo, which you should have set up as a remote.
$ git push <remote>  master
# Make the updated `master` the new base of the branch you made the changes on,
# which involves reapplying all the commits made to that branch.  Without the
# `--ignore-date` flag, git rebase changes the date on all the commits to the
# current date.
$ git checkout <branch name>
$ git rebase master --ignore-date
# Restore the changes you stashed earlier.
$ git stash pop

When you call rebase, you may run into some merge conflicts. Luke Champine's 'How to into git and GitHub' has more details (and many useful tricks).

Once the branch you want to test is up to date, you're ready to run some tests.

Testing the entire build

The make test command runs all tests (functions starting with Test in _test.go files) for each package, setting off a panic for any test that runs longer than 5s. For verbose output, run make test-v (which panics after 15s instead of 5s). Finally, make test-long has verbose output, only panics when a test takes 5 minutes, and also cleans up your code using gofmt and golint. You should run make test-long before each pull request.

Run make cover to run all tests for each package and generate color-coded .html visualizations of test coverage by function for each source file. Open cover/<module>.html in a browser to inspect a module's test coverage. For example, here's part of the html file generated for the persist package:

Screenshot

Meanwhile, make bench will call gofmt on all packages, then run all benchmarks (functions starting with Benchmark in _test.go files).

Testing a particular package or function

To run tests for just a certain package, run make test pkgs=./<package>. To run a certain test function, run make test pkgs=./<package> run=<function>. The same goes for make test-long, make cover and make bench.

For example, running test-long on the package persist produces this output:

$ make test-long pkgs=./persist
rm -rf release doc/whitepaper.aux doc/whitepaper.log doc/whitepaper.pdf
gofmt -s -l -w ./persist
go install ./persist
go vet ./persist
go test -v -race -tags='testing debug' -timeout=300s ./persist -run=Test
=== RUN   TestOpenDatabase
--- PASS: TestOpenDatabase (0.42s)
=== RUN   TestSaveLoad
--- PASS: TestSaveLoad (0.00s)
=== RUN   TestSaveLoadFile
--- PASS: TestSaveLoadFile (0.01s)
=== RUN   TestSaveLoadFileSync
--- PASS: TestSaveLoadFileSync (0.00s)
=== RUN   TestLogger
--- PASS: TestLogger (0.00s)
=== RUN   TestLoggerCritical
--- PASS: TestLoggerCritical (0.00s)
=== RUN   TestIntegrationRandomSuffix
--- PASS: TestIntegrationRandomSuffix (0.01s)
=== RUN   TestAbsolutePathSafeFile
--- PASS: TestAbsolutePathSafeFile (0.00s)
=== RUN   TestRelativePathSafeFile
--- PASS: TestRelativePathSafeFile (0.00s)
PASS
ok  	go.sia.tech/siad/persist	1.485s
$

Writing new tests for Sia

When you run make cover, you'll notice that many files have pretty low coverage. We're working on fixing that, but we could use your help.

A few guidelines

  • The test functions for filename.go should go in filename_test.go in the same directory and package.
  • A test function name should start with Test and clearly convey what is being tested.
  • You should declare function-specific variables and constants locally (inside the test function) instead of globally (outside the test function). That holds in general, not just for tests.
  • As always, code should adhere to the standards and conventions laid out in doc/Developers.md.
  • If a test interacts with the disk it should be skipped during the short tests by including the following at the beginning of the test: 
    if testing.Short() {
    t.SkipNow()
}

Basic test format

Suppose we'd like to test the Bar method belonging to type Foo.

// TestFoo checks that the Bar method on type Foo responds correctly to a normal
// input and returns the expected error when given a bad input.
func TestFoo(t *testing.T) {
	foo, err := NewFoo()
	if err != nil {
	// If NewFoo failed, we can't continue testing.
	t.Fatal(err)
	}

	// Try a normal input; should succeed.
	err := foo.Bar(3)
	if err != nil {
		// Report the error, but don't abort the test.
		t.Error(err)
	}

	// Try a bad input; should return an error.
	// NOTE: Always prefer to compare to a specific error, rather than 
	// err == nil
	err = Foo.Bar(0)
	if !errors.Contains(err, errDivideByZero){
		t.Errorf("expected errDivideByZero, got %v", err)
	}
}

Table-driven tests in Go

If you're looking to test a bunch of inputs, write a table-driven test with a slice of anonymous structs. For example, see TestParseFileSize in siac/parse_test.go:

func TestParseFilesize(t *testing.T) {
	// Define a table of test cases in the form of a slice of anonymous structs.
	tests := []struct {
		in, out string
		err     error
	}{
		{"1b", "1", nil},
		{"1KB", "1000", nil},
		{"1MB", "1000000", nil},
		{"1GB", "1000000000", nil},
		{"1TB", "1000000000000", nil},
		{"1KiB", "1024", nil},
		{"1MiB", "1048576", nil},
		{"1GiB", "1073741824", nil},
		{"1TiB", "1099511627776", nil},
		{"", "", errUnableToParseSize},
		{"123", "123", nil},
		{"123TB", "123000000000000", nil},
		{"123GiB", "132070244352", nil},
		{"123BiB", "", errUnableToParseSize},
		{"GB", "", errUnableToParseSize},
		{"123G", "", errUnableToParseSize},
		{"123B99", "", errUnableToParseSize},
		{"12A3456", "", errUnableToParseSize},
		{"1.23KB", "1230", nil},
		{"1.234KB", "1234", nil},
		{"1.2345KB", "1234", nil},
	}
	// Loop through the table of test cases to make sure ParseFileSize returns 
	// the expected output and error for each.
	for _, test := range tests {
		res, err := parseFilesize(test.in)
		if res != test.out || !errors.Contains(err, test.err){
			t.Errorf("parseFilesize(%v): expected %v %v, got %v %v", test.in, test.out, test.err, res, err)
		}
	}
}

Siatest Package

When deciding what tests to write for Sia, you should consider whether the best test is a unit test or a siatest. Ideally both a unit test and a siatest is written for new code. A unit test should explicitly test the new code functionality and a siatest should test the integration of the code and how it affects the rest of the platform. All the tests in the siatest package should use the API to execute the test. The idea is that everything that is trying to be tested should be able to be controlled and verified through the API. If you are looking for a place to start, there are many examples of older tests that could be upgraded to siatests.

Questions?

Read these if you haven't already:

Some other useful resources, some of which have been linked to already:

And feel free to ask questions on the #core-dev channel on the Sia Discord. Odds are, someone else is wondering the same thing.