Re: [Haskell-cafe] Re: `Expect'-like lazy reading/Parsec matching on TCP sockets
Hi Scott, On Mon, Apr 09, 2007 at 10:03:55AM -0600, Scott Bell wrote: Have you got a complete (but preferably small) program showing the problem? Great example, thanks! Sorry for the delay in tracking it down. main :: IO () main = do (_, h, _, p) - runInteractiveCommand telnet nyx.nyx.net t - hGetContentsTimeout h 15000 print t terminateProcess p The input handle is being garbage collected and closed, so telnet is exiting. Try: main :: IO () main = do (hin, h, _, p) - runInteractiveCommand telnet nyx.nyx.net t - hGetContentsTimeout h 15000 print t hClose hin terminateProcess p Note that you can't do either the hClose or terminateProcess before you have forced the whole string (which print does here). You might prefer to pass hin and p to hGetContentsTimeout, and have it close/terminate them just before the return . hGetContentsTimeout :: Handle - Int - IO String hGetContentsTimeout h t = do hSetBuffering stdin NoBuffering ready - hWaitForInput h t; eof - hIsEOF h You'll also need to remove the hIsEOF call from your code, or having decided that nothing is ready it will then block, waiting to see if there is an end of file or not. Thanks Ian ___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
Re: [Haskell-cafe] Re: `Expect'-like lazy reading/Parsec matching on TCP sockets
Have you got a complete (but preferably small) program showing the problem? Ian, Here is the source and behavior that I'm seeing (Linux x86, under both 6.6 and 6.7-20070404: module Main where import System.IO import System.IO.Unsafe import System.Process import Text.ParserCombinators.Parsec main :: IO () main = do (_, h, _, p) - runInteractiveCommand telnet nyx.nyx.net t - hGetContentsTimeout h 15000 print t terminateProcess p hGetContentsTimeout :: Handle - Int - IO String hGetContentsTimeout h t = do hSetBuffering stdin NoBuffering ready - hWaitForInput h t; eof - hIsEOF h if ((not ready) || eof) then return [] else do c - hGetChar h s - unsafeInterleaveIO (hGetContentsTimeout h t) return (c:s) -- Behavior with threaded RTS, string is returned early because of EOF: [EMAIL PROTECTED]:~/src/remote$ ghc --make Remote.hs -o remote -threaded [1 of 1] Compiling Main ( Remote.hs, Remote.o ) Linking remote ... [EMAIL PROTECTED]:~/src/remote$ ./remote Trying 206.124.29.1...\nConnected to nyx.nyx.net.\nEscape character is '^]'.\n -- Behavior with non-threaded RTS, proper timeout is observed: [EMAIL PROTECTED]:~/src/remote$ ./remote Trying 206.124.29.1...\nConnected to nyx.nyx.net.\nEscape character is '^]'.\n\n\n Welcome to Nyx, The Spirit of the Night\n(303) 409-1401\n nyx.nyx.net -- 206.124.29.1\n nyx10.nyx.net -- 206.124.29.2\n\n Free Public Internet Access\n\n ===\n New user? Login as new\n ===\n\n (If you get timed out, try later. Nyx would be too slow to use.)\n\n\n\n\r\n\r\nSunOS UNIX (nyx)\r\n\r\r\n\rlogin: ___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
[Haskell-cafe] Re: `Expect'-like lazy reading/Parsec matching on TCP sockets
you should write compiler version and OS for such problems. Of course, this was 6.7-20070404 on Linux x86. I haven't had a change to test it on my 6.6 build. I/O in Windows threaded RTS was fixed after initial 6.6 release and afaik hWaitForInput should work better now. alternatively, you can try with hand-made timeout solution I'm not sure what the status is on hWaitForInput with the snapshot that I am using -- any other ideas what I could try to further isolate the problem? - Scott Bell ___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
Re: [Haskell-cafe] Re: `Expect'-like lazy reading/Parsec matching on TCP sockets
On Fri, Apr 06, 2007 at 01:44:01PM -0600, Scott Bell wrote: Ooops! It seems that this doesn't behave well with a -threaded RTS. I get an EOF on handles that I know for a fact shouldn't be receiving them. It still works well without -threaded, but does anyone know why I'm getting this behavior? hGetContentsTimeout :: Handle - Int - IO String hGetContentsTimeout h t = do hSetBuffering stdin NoBuffering ready - hWaitForInput h t if (not ready) then return [] else do c - hGetChar h s - unsafeInterleaveIO (hGetContentsTimeout h t) return (c:s) (I did add EOF checking, but all that did was return the end of the list earlier than I wanted) Have you got a complete (but preferably small) program showing the problem? Thanks Ian ___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
[Haskell-cafe] Re: `Expect'-like lazy reading/Parsec matching on TCP sockets
Ooops! It seems that this doesn't behave well with a -threaded RTS. I get an EOF on handles that I know for a fact shouldn't be receiving them. It still works well without -threaded, but does anyone know why I'm getting this behavior? hGetContentsTimeout :: Handle - Int - IO String hGetContentsTimeout h t = do hSetBuffering stdin NoBuffering ready - hWaitForInput h t if (not ready) then return [] else do c - hGetChar h s - unsafeInterleaveIO (hGetContentsTimeout h t) return (c:s) (I did add EOF checking, but all that did was return the end of the list earlier than I wanted)___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
Re: [Haskell-cafe] Re: `Expect'-like lazy reading/Parsec matching on TCP sockets
Hello Scott, Friday, April 6, 2007, 11:44:01 PM, you wrote: you should write compiler version and OS for such problems. if it's 6.6 and Win, try to upgrade to current STABLE build: smth like http://www.haskell.org/ghc/dist/stable/dist/ghc-6.6.20061207-i386-unknown-mingw32.tar.gz but newer (this one is dec 12 build) I/O in Windows threaded RTS was fixed after initial 6.6 release and afaik hWaitForInput should work better now. alternatively, you can try with hand-made timeout solution Ooops! It seems that this doesn't behave well with a -threaded RTS. I get an EOF on handles that I know for a fact shouldn't be receiving them. It still works well without -threaded, but does anyone know why I'm getting this behavior? hGetContentsTimeout :: Handle - Int - IO String hGetContentsTimeout h t = do hSetBuffering stdin NoBuffering ready - hWaitForInput h t if (not ready) then return [] else do c - hGetChar h s - unsafeInterleaveIO (hGetContentsTimeout h t) return (c:s) (I did add EOF checking, but all that did was return the end of the list earlier than I wanted) -- Best regards, Bulatmailto:[EMAIL PROTECTED] ___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
