Reputation: 1401
I noticed some unexpected behavior when using Mutex so I decided to write some test code.
package main
import (
"fmt"
"sync"
)
type Box struct {
Stuff int
mu sync.Mutex
}
var box Box
func init() {
box = Box{}
box.mu.Lock()
}
func (box *Box) add() {
fmt.Println("add()")
box.mu.Unlock()
box.Stuff++
box.mu.Lock()
}
func (box *Box) display() {
fmt.Print("display(): ")
box.mu.Unlock()
fmt.Println(box.Stuff)
box.mu.Lock()
}
func add2() {
fmt.Println("add2()")
box.Stuff++ //increments but should not, error expected
}
func display2() {
fmt.Print("display2(): ")
fmt.Println(box.Stuff) //displays but should not, error expected
}
func main() {
box.display()
box.add()
box.display()
display2()
add2()
box.display()
}
The functions without pointer receivers can add & access the protected struct. Running the code.
display(): 0
add()
display(): 1
display2(): 1
add2()
display(): 2
I would expect the two functions which have the name appended with a 2 to fail. They do not.
Any idea?
Upvotes: 0
Views: 71
Reputation: 6084
You might want to write this a bit different:
You start by locking, which is not the most common way to use a lock: Everything is locked unless you unlock it.
A more common way is:
package main
import (
"fmt"
"sync"
)
type Box struct {
Stuff int
mu sync.Mutex
}
var box Box
func init() {
box = Box{}
}
func (box *Box) add() {
fmt.Println("add()")
box.mu.Lock()
box.Stuff++
box.mu.Unlock()
}
...
Scoping your locks to the moment of change and not to the complete app.
Upvotes: 3