Beefy Boxes and Bandwidth Generously Provided by pair Networks RobOMonk
Perl: the Markov chain saw
 
PerlMonks  

Re^2: Locked threads and tcp timeouts

by BrowserUk (Pope)
on Feb 16, 2012 at 17:05 UTC ( #954282=note: print w/ replies, xml ) Need Help??


in reply to Re: Locked threads and tcp timeouts
in thread Locked threads and tcp timeouts

This will not work if the thread is in a blocking IO read state. Ie. The OPs problem.


Comment on Re^2: Locked threads and tcp timeouts
Re^3: Locked threads and tcp timeouts
by nikosv (Hermit) on Feb 19, 2012 at 15:42 UTC
    would a more drastic approach work ? like getting the underlying Win32 handle ($thr->_handle()) of the worker thread (since threads in Windows are created with CreateThread after all), and kill the thread by passing the handle to the TerminateThread api call ?

    a note to the OP, when killing threads associated with calls to the dbms i.e sql execution, might lead to locked resources i.e tables or rows since there is no proper clean up

      There are two problems with the idea.

      1. I've never managed to work out how to use the value returned from _handle().

        It appears to be returned in the IV slot of the SV:

        SV = IV(0x3afb470) at 0x3afb478 REFCNT = 1 FLAGS = (PADMY,IOK,pIOK) IV = 94514760

        but all my attempts to use it from XS in calls to the OS result in The handle is invalid.

      2. Terminating threads is a desperate measure as it does not allow either the thread itself, the OS, or Perl to clean up after themselves, leading to a massive memory leak. Not to mention the potential for all sorts of nasties with dangling pointers etc.

        If you bypass the first problem by calling an XS sub that gets the current thread handle from the OS -- which isn't very useful as it can only be called from within the thread to be killed -- and run a loop that creates threads and has them self terminate, then you see the scope of the massive memory leak it creates:

        #! perl -slw use strict; use Inline C => Config => BUILD_NOISY => 1; use Inline C => <<'END_C', NAME => 'TerminateThreadTest', CLEAN_AFTER +_BUILD => 0; int killThread( SV *t, SV *ec ) { printf( "t:%p ec:%d\n", SvIV( t ), SvIV( ec ) ); return TerminateThread( SvIV( t ), SvIV( ec ) ); } void hariKari( SV *exitCode ) { TerminateThread( GetCurrentThread(), SvIV( exitCode ) ); } END_C use Devel::Peek; use threads; for( 1 .. 100 ) { my $t = async { sleep 5; hariKari( 12345 ); }; sleep 10; }

      On my system, the above script leaks around 2MB for every thread killed. Whilst that could be reduce by spawning early and trimming the stack etc., it will never be zero.


      With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday'
      Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
      "Science is about questioning the status quo. Questioning authority".
      In the absence of evidence, opinion is indistinguishable from prejudice.

      The start of some sanity?

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://954282]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others chilling in the Monastery: (6)
As of 2014-04-19 05:14 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    April first is:







    Results (478 votes), past polls