chore: libwaku - better error handling and better waku thread destroy handling (#3167)

apps/networkmonitor/networkmonitor_metrics.nim

@@ -37,8 +37,7 @@ declarePublicGauge networkmonitor_peer_user_agents,
 
 declarePublicHistogram networkmonitor_peer_ping,
   "Histogram tracking ping durations for discovered peers",
-  buckets =
+  buckets = [10.0, 20.0, 50.0, 100.0, 200.0, 300.0, 500.0, 800.0, 1000.0, 2000.0, Inf]
-    [10.0, 20.0, 50.0, 100.0, 200.0, 300.0, 500.0, 800.0, 1000.0, 2000.0, Inf]
 
 declarePublicGauge networkmonitor_peer_count,
   "Number of discovered peers", labels = ["connected"]

library/libwaku.nim

@@ -168,7 +168,7 @@ proc waku_destroy(
 ): cint {.dynlib, exportc.} =
   checkLibwakuParams(ctx, callback, userData)
 
-  waku_thread.stopWakuThread(ctx).handleRes(callback, userData)
+  waku_thread.destroyWakuThread(ctx).handleRes(callback, userData)
 
 proc waku_version(
     ctx: ptr WakuContext, callback: WakuCallBack, userData: pointer

library/waku_thread/waku_thread.nim

@@ -18,36 +18,45 @@ type WakuContext* = object
   userData*: pointer
   eventCallback*: pointer
   eventUserdata*: pointer
+  running: Atomic[bool] # To control when the thread is running
 
 const git_version* {.strdefine.} = "n/a"
 const versionString = "version / git commit hash: " & waku.git_version
 
-# To control when the thread is running
-# TODO: this should be part of the context so multiple instances can be executed
-var running: Atomic[bool]
-
 proc runWaku(ctx: ptr WakuContext) {.async.} =
   ## This is the worker body. This runs the Waku node
   ## and attends library user requests (stop, connect_to, etc.)
-  info "Starting Waku", version = versionString
 
   var waku: Waku
 
-  while running.load == true:
+  while true:
     await ctx.reqSignal.wait()
 
-    # Trying to get a request from the libwaku main thread
+    if ctx.running.load == false:
+      break
+
+    ## Trying to get a request from the libwaku requestor thread
     var request: ptr InterThreadRequest
     let recvOk = ctx.reqChannel.tryRecv(request)
-    if recvOk == true:
+    if not recvOk:
-      let resultResponse = waitFor InterThreadRequest.process(request, addr waku)
+      error "waku thread could not receive a request"
+      continue
 
-      ## Converting a `Result` into a thread-safe transferable response type
+    ## Handle the request
-      let threadSafeResp = InterThreadResponse.createShared(resultResponse)
+    let resultResponse = waitFor InterThreadRequest.process(request, addr waku)
 
-      ## The error-handling is performed in the main thread
+    ## Converting a `Result` into a thread-safe transferable response type
-      discard ctx.respChannel.trySend(threadSafeResp)
+    let threadSafeResp = InterThreadResponse.createShared(resultResponse)
-      discard ctx.respSignal.fireSync()
+
+    ## Send the response back to the thread that sent the request
+    let sentOk = ctx.respChannel.trySend(threadSafeResp)
+    if not sentOk:
+      error "could not send a request to the requester thread",
+        original_request = $request[]
+
+    let fireRes = ctx.respSignal.fireSync()
+    if fireRes.isErr():
+      error "could not fireSync back to requester thread", error = fireRes.error
 
 proc run(ctx: ptr WakuContext) {.thread.} =
   ## Launch waku worker
@@ -62,7 +71,7 @@ proc createWakuThread*(): Result[ptr WakuContext, string] =
   ctx.respSignal = ThreadSignalPtr.new().valueOr:
     return err("couldn't create respSignal ThreadSignalPtr")
 
-  running.store(true)
+  ctx.running.store(true)
 
   try:
     createThread(ctx.thread, run, ctx)
@@ -74,15 +83,19 @@ proc createWakuThread*(): Result[ptr WakuContext, string] =
 
   return ok(ctx)
 
-proc stopWakuThread*(ctx: ptr WakuContext): Result[void, string] =
+proc destroyWakuThread*(ctx: ptr WakuContext): Result[void, string] =
-  running.store(false)
+  ctx.running.store(false)
-  let fireRes = ctx.reqSignal.fireSync()
+
-  if fireRes.isErr():
+  let signaledOnTime = ctx.reqSignal.fireSync().valueOr:
-    return err("error in stopWakuThread: " & $fireRes.error)
+    return err("error in destroyWakuThread: " & $error)
-  discard ctx.reqSignal.close()
+  if not signaledOnTime:
-  discard ctx.respSignal.close()
+    return err("failed to signal reqSignal on time in destroyWakuThread")
+
   joinThread(ctx.thread)
+  ?ctx.reqSignal.close()
+  ?ctx.respSignal.close()
   freeShared(ctx)
+
   return ok()
 
 proc sendRequestToWakuThread*(