RingBuffer with a size of 1 should not be allowed due to an insert bug, so increment a size of one, by one.

[hedisam]

RingBuffer with a size of 1 will have a mask of 0 which uncovers a bug in the algorithm since the mask value must only contain ones (left padded with zeros) if it's represented in its binary format (e.g. '0111', '0011 1111', etc.). So we should not allow having RingBuffer queues with a size of 1. We can simply fix this by incrementing the size inside the RingBuffer's init method.

With a full RingBuffer of size 1, insert operations replace the existing item in the queue and then puts the RingBuffer into an invalid state.

[aviary-wf]

Security Insights

No security relevant content was detected by automated scans.

Action Items

• Review PR for security impact; comment "security review required" if needed or unsure
• Verify aviary.yaml coverage of security relevant code

---

Questions or Comments? Reach out on Slack: #support-infosec.

[hedisam]

You can see the bug in action by running this unit test:

func TestRingInsertSizeOne(t *testing.T) {
	rb := NewRingBuffer(1)
	assert.Equal(t, uint64(1), rb.Cap())

	err := rb.Put("Hello")
	if !assert.Nil(t, err) {
		return
	}

	ok, err := rb.Offer("Hello, Again.")
	if !assert.Nil(t, err) {
		return
	}
	assert.False(t, ok)

        // rb.Get() blocks the routine since our RingBuffer's state becomes invalid 
	result, err := rb.Poll(time.Millisecond * 100)
	if !assert.Nil(t, err) {
		return
	}
	if !assert.NotNil(t, result) {
		return
	}
	assert.Equal(t, "Hello", result)
}