Atlas › Test
TestTryAcquire_LowAfterWaiting
Exact test identity: go.temporal.io/server/common/locks/TestPrioritySemaphoreSuite/TestTryAcquire_LowAfterWaiting
- Package
go.temporal.io/server/common/locks- Suite / test hierarchy
TestPrioritySemaphoreSuite/TestTryAcquire_LowAfterWaiting- Test
TestTryAcquire_LowAfterWaiting- Introduced at
- TestTryAcquire_LowAfterWaiting, TestTryAcquire_HighAfterWaiting Frontier kind: Test frontier
- Covered ranges
- 20
- Covered lines
- 62
- Covered files
- 1
Co-introduced tests
1 other test enter at the same concept.
Covered source
Expand a file to inspect source; the > gutter marks covered lines.
go.temporal.io/server/common/locks/priority_semaphore_impl.go 62 covered LOC · 20 ranges
Open complete file
67
// maximum combined weight for concurrent access, capable of handling multiple priority levels.
68
// Most of the logic is taken directly from golang's semaphore.Weighted.
70
>
waitLists := make([]*list.List, NumPriorities)
71
>
for i := range waitLists {
72
>
waitLists[i] = list.New()
73
>
}
74
>
return &PrioritySemaphoreImpl{
75
>
size: n,
76
>
waitLists: waitLists,
77
>
}
78
}
79
81
// are available or ctx is done. On success, returns nil. On failure, returns
82
// ctx.Err() and leaves the semaphore unchanged.
84
>
if priority >= NumPriorities {
85
// nolint:forbidigo
86
panic(fmt.Sprintf("semaphore: invalid priority %v, priority must be less than %v", priority, NumPriorities))
87
}
88
90
>
91
>
s.mu.Lock()
92
>
select {
93
case <-done:
94
// ctx becoming done has "happened before" acquiring the semaphore,
98
s.mu.Unlock()
99
return ctx.Err()
101
}
102
// Check if acquisition can proceed without waiting
104
// Since we hold s.mu and haven't synchronized since checking done, if
105
// ctx becomes done before we return here, it becoming done must have
111
}
112
114
s.mu.Unlock()
115
return ErrRequestTooLarge
116
}
117
119
>
w := waiter{n: n, ready: ready}
120
>
elem := s.waitLists[priority].PushBack(w)
121
>
s.mu.Unlock()
122
>
123
>
select {
124
case <-done:
125
s.mu.Lock()
141
return ctx.Err()
142
144
>
// Acquired the semaphore. Check that ctx isn't already done.
145
>
// We check the done channel instead of calling ctx.Err because we
146
>
// already have the channel, and ctx.Err is O(n) with the nesting
147
>
// depth of ctx.
148
>
select {
149
case <-done:
150
s.Release(n)
151
return ctx.Err()
153
}
155
}
156
}
158
// TryAcquire acquires the semaphore with a weight of n without blocking.
159
// On success, returns true. On failure, returns false and leaves the semaphore unchanged.
161
>
if priority >= NumPriorities {
162
// nolint:forbidigo
163
panic(fmt.Sprintf("semaphore: invalid priority %v, priority must be less than %v", priority, NumPriorities))
164
}
165
167
>
defer s.mu.Unlock()
168
>
if s.size-s.cur >= n && s.noWaiters(priority) {
169
>
s.cur += n
170
>
return true
171
>
}
172
return false
173
}
174
176
>
s.mu.Lock()
177
>
defer s.mu.Unlock()
178
>
s.cur -= n
179
>
if s.cur < 0 {
180
s.mu.Unlock()
181
panic("semaphore: released more than held")
182
}
184
}
185
187
>
for _, l := range s.waitLists {
188
>
for {
189
>
next := l.Front()
190
>
if next == nil {
191
>
break // No more waiters blocked.
192
}
193
195
>
if !ok {
196
panic("semaphore: failed to cast waiter")
197
}
199
// Not enough tokens for the next waiter. We could keep going (to try to
200
// find a waiter with a smaller request), but under load that could cause
211
}
212
214
>
l.Remove(next)
215
>
close(w.ready)
216
}
217
}
219
220
// noWaiters returns if there is no waiter that has priority higher or equal to lowestPriority.
222
>
for _, l := range s.waitLists[:lowestPriority+1] {
223
>
if l.Len() > 0 {
224
return false
225
}
226
}
228
}