a94af5c · distributed-process

git.iko.soy/distributed-process

Break feedback loop when tracing messages to the console.

If the user reregisters the "logger" with a custom process which uses 'send' or other primitives which are traced, this primitives would produce tracing messages which would cause the primitives to be called again. This bug only affects tracing messages to the console with DISTRIBUTED_PROCESS_TRACE_CONSOLE=y DISTRIBUTED_PROCESS_TRACE_FLAGS=r

commita94af5c07a9adbfbc4172452a39b2360e8593684 authorFacundo Domínguez <facundo.dominguez@tweag.io> dateThu 5 Feb 2015, 15:43 (12 years ago) parentdb5aa53

2 files changed+7−2

compare with master

unifiedsplit

src/Control/Distributed/Process/Management/Internal/Trace/Tracer.hs+1−1 src/Control/Distributed/Process/Node.hs+6−1

src/Control/Distributed/Process/Management/Internal/Trace/Tracer.hs+1−1

@@ -146,7 +146,7 @@ systemLoggerTracer = do

146146 emptyPid <- return $ (nullProcessId (localNodeId node))

147147 traceMsg <- return $ NCMsg {

148148 ctrlMsgSender = ProcessIdentifier (emptyPid)

149− , ctrlMsgSignal = (NamedSend "logger"

149+ , ctrlMsgSignal = (NamedSend "trace.logger"

150150 (createUnencodedMessage msg))

151151 }

152152 liftIO $ writeChan (localCtrlChan node) traceMsg

src/Control/Distributed/Process/Node.hs+6−1

@@ -295,7 +295,12 @@ startServiceProcesses node = do

295295 tableCoordinatorPid <- fork $ Table.startTableCoordinator fork

296296 runProcess node $ register Table.mxTableCoordinator tableCoordinatorPid

297297 logger <- forkProcess node loop

298− runProcess node $ register "logger" logger

298+ runProcess node $ do

299+ register "logger" logger

300+ -- The trace.logger is used for tracing to the console to avoid feedback

301+ -- loops during tracing if the user reregisters the "logger" with a custom

302+ -- process which uses 'send' or other primitives which are traced.

303+ register "trace.logger" logger

299304 where

300305 fork = forkProcess node

301306