diff options
| -rw-r--r-- | utils.go | 3 | ||||
| -rw-r--r-- | ws.go | 19 |
2 files changed, 21 insertions, 1 deletions
@@ -23,11 +23,14 @@ package main import ( "crypto/rand" "encoding/base64" + "errors" "fmt" "log" "net/http" ) +var errUnexpectedRace = errors.New("unexpected race condition") + /* * Write a string to a http.ResponseWriter, setting the Content-Type and status * code. @@ -217,13 +217,30 @@ var ( ) func setupChanPool() error { - chanPoolLock.Lock() + /* + * It would be unusual for this function to run concurrently with + * anything else that modifies chanPool, so we fail when the lock is + * unsuccessful. + */ + r := chanPoolLock.TryLock() + if !r { + return fmt.Errorf("cannot set up chanPool: %w", errUnexpectedRace) + } defer chanPoolLock.Unlock() chanPool = make(map[string](*chan string)) return nil } func propagate(msg string) { + /* + * It is not a mistake that we acquire a read lock instead of a write + * lock here. Channels provide synchronization, and other than using + * the channels, we are simply iterating through chanPoolLock. This is + * unsafe when chanPoolLock's structure is being modified, such as + * when a channel is being added or deleted from the pool; but it's + * fine if other goroutines are simply indexing it and using the + * channels. + */ chanPoolLock.RLock() defer chanPoolLock.RUnlock() for k, v := range chanPool { |