Commit 4f4f12e5 authored by Simon Marlow's avatar Simon Marlow
Browse files

attempt to work around restrictions with fork() & pthreads

In the child process, call exec() directly instead of using
System.Cmd.system, which involves another fork()/exec() and a
non-blocking wait.  The problem is that in a forked child of a
threaded process, it isn't safe to do much except exec() according to
POSIX.  In fact calling pthread_create() in the child causes the
pthread library to fail with an error on FreeBSD.
parent 4e705b5c
......@@ -12,7 +12,7 @@ import System.Process
import Control.Monad (when)
#if !defined(mingw32_HOST_OS)
import System.Process.Internals (mkProcessHandle)
import System.Posix.Process (forkProcess, createSession)
import System.Posix.Process (forkProcess, createSession, executeFile)
import System.Posix.Signals (installHandler, Handler(Catch),
signalProcessGroup, sigINT, sigTERM, sigKILL )
......@@ -30,10 +30,7 @@ main = do
forkIO (do threadDelay (read secs * 1000000)
putMVar m Nothing
forkIO (do try (do pid <- forkProcess $ do
r <- system cmd
exitWith r
forkIO (do try (do pid <- systemSession cmd
ph <- mkProcessHandle pid
putMVar mp (pid,ph)
r <- waitForProcess ph
......@@ -51,6 +48,16 @@ main = do
_other -> do hPutStrLn stderr "timeout: bad arguments"
exitWith (ExitFailure 1)
systemSession cmd =
forkProcess $ do
executeFile "/bin/sh" False ["-c", cmd] Nothing
-- need to use exec() directly here, rather than something like
-- System.Process.system, because we are in a forked child and some
-- pthread libraries get all upset if you start doing certain
-- things in a forked child of a pthread process, such as forking
-- more threads.
killProcess pid ph = do
try (signalProcessGroup sigTERM pid)
checkReallyDead 10
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment