more experiments

tests/exampleGcFailures/exFailureSeqSeq.nim

@@ -29,13 +29,14 @@ proc toSeqDataPtr*[T](data: seq[T]): SeqDataPtr[T] =
       data: cast[ptr UncheckedArray[T]](unsafeAddr(data[0])), size: data.len()
     )
 
-proc worker(data: ptr seq[seq[char]], sig: ThreadSignalPtr) =
-  os.sleep(10)
+proc worker(data: seq[seq[char]], sig: ThreadSignalPtr) =
+  # os.sleep(10)
   echo "running worker: "
-  echo "worker: ", data[]
-  for i, d in data[]:
-    for j, c in d:
-      data[i][j] = char(c.uint8 + 10)
+  echo "worker: ", data
+  # for i, d in data:
+  #   for j, c in d:
+  #     d[j] = char(c.uint8 + 10)
+  GC_fullCollect()
   discard sig.fireSync()
 
 proc runTest(tp: TaskPool, sig: ThreadSignalPtr) {.async.} =
@@ -45,27 +46,28 @@ proc runTest(tp: TaskPool, sig: ThreadSignalPtr) {.async.} =
   var data = @[obj1, obj2]
 
   # echo "spawn worker"
-  tp.spawn worker(addr data, sig)
+  tp.spawn worker(data, sig)
 
   ## adding fut.wait(100.milliseconds) creates memory issue
-  await wait(sig).wait(1.milliseconds)
+  await wait(sig)
   ## just doing the wait is fine:
   # await wait(sig)
+  echo "data: ", data
 
 proc runTests(tp: TaskPool, sig: ThreadSignalPtr) {.async.} =
-  for i in 1..30_000:
+  for i in 1..10_000:
     try:
       await runTest(tp, sig)
     except AsyncTimeoutError:
       # os.sleep(1)
       # echo "looping..."
-      # GC_fullCollect()
       discard
+    GC_fullCollect()
 
 suite "async tests":
-  var tp = Taskpool.new(num_threads = 2) # Default to the number of hardware threads.
+  var tp = Taskpool.new(num_threads = 8) # Default to the number of hardware threads.
   let sig = ThreadSignalPtr.new().get()
 
   asyncTest "test":
     await runTests(tp, sig)
-    os.sleep(10_000)
+    # os.sleep(10_000)

tests/exampleGcFailures/exFailureSeqSeqCursor.nim

@@ -0,0 +1,63 @@
+import std/os
+import std/sequtils
+import chronos
+import chronos/threadsync
+import chronos/unittest2/asynctests
+import taskpools
+
+## create a probablistically likely failure of 
+## using sequence memory from another thread
+## with refc. 
+## 
+## However, unlike `exFailure.nim`, this can take
+## a while to run.
+## 
+## It may not always produce an error either, but
+## generally does so in a few seconds of running.
+## 
+
+type
+  SeqCursor* = object
+    data* {.cursor.}: seq[seq[char]]
+
+proc worker(data: SeqCursor, sig: ThreadSignalPtr) =
+  os.sleep(10)
+  echo "running worker: "
+  echo "worker: ", data.data
+  # for i, d in data.data:
+  #   for j, c in d:
+  #     data.data[i][j] = char(c.uint8 + 10)
+  discard sig.fireSync()
+
+proc runTest(tp: TaskPool, sig: ThreadSignalPtr) {.async.} =
+  ## init
+  var obj1 = "hello world!".toSeq()
+  var obj2 = "goodbye denver!".toSeq()
+  var data = @[obj1, obj2]
+  var cur = SeqCursor(data: data)
+
+  # echo "spawn worker"
+  tp.spawn worker(cur, sig)
+
+  ## adding fut.wait(100.milliseconds) creates memory issue
+  await wait(sig)
+  ## just doing the wait is fine:
+  # await wait(sig)
+
+proc runTests(tp: TaskPool, sig: ThreadSignalPtr) {.async.} =
+  for i in 1..3_000:
+    try:
+      await runTest(tp, sig)
+    except AsyncTimeoutError:
+      # os.sleep(1)
+      # echo "looping..."
+      # GC_fullCollect()
+      discard
+
+suite "async tests":
+  var tp = Taskpool.new(num_threads = 2) # Default to the number of hardware threads.
+  let sig = ThreadSignalPtr.new().get()
+
+  asyncTest "test":
+    await runTests(tp, sig)
+    os.sleep(10_000)

tests/exampleGcFailures/test.nim

@@ -0,0 +1,25 @@
+
+when false:
+  type AnObject* = object of RootObj
+    value*: int
+
+  proc mutate(a: sink AnObject) =
+    a.value = 1
+
+  var obj = AnObject(value: 42)
+  mutate(obj)
+  doAssert obj.value == 42
+
+else:
+  type AnObject = object of RootObj
+    value*: int
+
+  proc `=destroy`(x: var AnObject) = 
+    echo "DEST"
+
+  proc mutate(a: sink AnObject) =
+    a.value = 1
+
+  var obj = AnObject(value: 42)
+  mutate(obj)
+  doAssert obj.value == 42