I think you will find it very difficult to control what errors are thrown and when by os.Stat
in a platform independent way from a unit test. If you really need a test for the path where an unknown error type is returned, your best bet might be to refactor your package code so that you can mock os.Stat
. Although you can't change the behavior of os.Stat
directly, by taking advantage of the fact that Go has first-class functions, you can use a bit of indirection to mock it with very minimal changes to your code. If your package code looks like this:
package mypackage
import "os"
...
if _, err := os.Stat(path); err != nil {
if os.IsNotExist(err) {
continue
}
return errors.File().AddDetails(err)
}
...
Try refactoring it to use a non-exported package-scope function variable which is assigned os.Stat
:
package mypackage
import "os"
var osStat = os.Stat
...
if _, err := osStat(path); err != nil {
if os.IsNotExist(err) {
continue
}
return errors.File().AddDetails(err)
}
...
Now, in your test code (which should be in the same package as the code under test), you can reassign osStat
to any function with the same signature in order to mock it:
package mypackage
import (
"os"
"testing"
)
func TestNotExistError(t *testing.T) {
osStat = func(string) (os.FileInfo, error) {
return nil, os.ErrNotExist
}
// in case other test functions depend on the unmocked behavior
defer func() {
osStat = os.Stat
}()
// rest of the test which triggers the codepath above
}
func TestOtherError(t *testing.T) {
osStat = func(string) (os.FileInfo, error) {
return nil, os.ErrInvalid
}
// in case other test functions depend on the unmocked behavior
defer func() {
osStat = os.Stat
}()
// rest of the test which triggers the codepath above
}
...