Unix.waitpid

Unix.waitpid

[Bauer, Christoph]

Hi,

my app should wait for some processes. This is done 
by my function wait, which gets a pid, a idle function (e.g. ignore)
+ its value and a delay time between polling:


let delay s = ignore (Unix.select [] [] [] s)

let wait ?(delay_time = 0.01) pid idle x =
  let rec wait' () =
    idle x;
    delay delay_time;
    let pid = try fst (Unix.waitpid [Unix.WNOHANG] pid) with _  -> 1 in
      if pid <> 0 then wait' ()
  in wait' ()

It works mostly, but in some rare, non-deterministic cases wait loops
forever.
In a seconds round it works without problems. I saw this behaviour on Linux
and Windows. Can anybody help me?

Thanks!

Christoph Bauer
Dipl. Inf.

LMS Deutschland GmbH
Luxemburgerstr. 7
D-67657 Kaiserslautern

T +49 631 303 22 152

mailto:Christoph.Bauer@lms-gmbh.de
http://www.lmsintl.com 

Re: [Caml-list] Unix.waitpid

[Oliver Bandel]

On Thu, Jun 30, 2005 at 02:09:51PM +0200, Bauer, Christoph wrote:
> Hi,
> 
> my app should wait for some processes. This is done 
> by my function wait, which gets a pid, a idle function (e.g. ignore)
> + its value and a delay time between polling:
> 
> 
> let delay s = ignore (Unix.select [] [] [] s)
> 
> let wait ?(delay_time = 0.01) pid idle x =
>   let rec wait' () =
>     idle x;
>     delay delay_time;
>     let pid = try fst (Unix.waitpid [Unix.WNOHANG] pid) with _  -> 1 in
>       if pid <> 0 then wait' ()
>   in wait' ()

Does you have a valid pid (as calling parameter)?
You may better use 0 or -1 in the call to Unix.waitpid.

Maybe WUNTRACED would also makes sense as flag.
Maybe better not testing with "_" in try-with.


P.S.: Documentation could be better.
I had to look in the C-API manpage to know what parameters
the Unix.aitpid awaits.
That the one arg is the pid is mentioned in the documentation, but there
is no function call example, to identify the second arg as the wpid.


Ciao,
   Oliver Bandel

AW: [Caml-list] Unix.waitpid

[Bauer, Christoph]

Sorry,

> let wait ?(delay_time = 0.01) pid idle x =
>   let rec wait' () =
>     idle x;
>     delay delay_time;
>     let pid = try fst (Unix.waitpid [Unix.WNOHANG] pid) with 
> _  -> 1 in
>       if pid <> 0 then wait' ()
>   in wait' ()
this code is wrong, my original code was:

let rec wait' () =
      idle x;
      delay delay_time;
     let wpid = try fst( Unix.waitpid [Unix.WNOHANG] pid) with _ -> pid in
       if pid <> wpid then wait' ()
    in wait' ()

Christoph Bauer

Re: [Caml-list] Unix.waitpid

[Oliver Bandel]

On Thu, Jun 30, 2005 at 02:31:06PM +0200, Bauer, Christoph wrote:
> Sorry,
> 
> > let wait ?(delay_time = 0.01) pid idle x =
> >   let rec wait' () =
> >     idle x;
> >     delay delay_time;
> >     let pid = try fst (Unix.waitpid [Unix.WNOHANG] pid) with 
> > _  -> 1 in
> >       if pid <> 0 then wait' ()
> >   in wait' ()
> this code is wrong, my original code was:
> 
> let rec wait' () =
>       idle x;
>       delay delay_time;
>      let wpid = try fst( Unix.waitpid [Unix.WNOHANG] pid) with _ -> pid in
>        if pid <> wpid then wait' ()
>     in wait' ()


Well, using the corresponding C-API-function,
you should better rename your arguments:

  wpid is the pid you wait for (the argument to Unix.waitpid)  and
   pid is the value you get from that function:


===== from the wait(2)/waitpid(2)-manpage =============
     pid_t
     waitpid(pid_t wpid, int *status, int options);
=======================================================


BTW: where do you set the "pid" initially?

And: if waitpid returns with a pid, then you can't wait on this
pid/process again. So, if you wait for one pid and get a different
pid as result of the function call, then you wait again.
But if you another time waits for a pid, which you alread caught
(got as a result once before), then you can wait infinite,
because that pid was caught by waitpud before... and you have to
wait until the pid is reused by the kernel or wait until next reboot. ;-)

So your concept seems to be buggy.

Ciao,
   Oliver Bandel