Version française
Home     About     Download     Resources     Contact us    
Browse thread
Interfacing C-code with ocaml and Exception handling
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: Christian Sternagel <christian.sternagel@u...>
Subject: Interfacing C-code with ocaml and Exception handling
I'm not sure whether this question was asked before (at least I didn't find any answer). Can it be, that raising an (OCaml)Exception during execution of
C-code (interfaced with OCaml) is simply ignored?

More concretely an example:

We have a module Timer that implements execution of functions given a certain timeout. Therefor we have the function

 Timer.run : float -> (unit -> 'a) -> 'a

where [Timer.run t f] starts a timer raising the signal SIG_ALRM after
[t] seconds and executes [f ()]. If [f] finishes in time the result of
[f ()] is returned. In the background a handler for SIG_ALRM is installed, that just raises the exception [Timer.Timeout]. The implementation of
[Timer.run] looks like this:

 let run t f =
  try
   start t;
   let result = f () in
   stop ();
   result
  with
   | e -> stop (); raise e
 ;;
 
and the handler assigned to SIG_ALRM like this

 let handler _ = raise Timeout;;

where [start] and [stop] take care of a [Unix.itimer]. The intended behaviour is that if [f ()] finishes within time the result is returned and the exception [Timeout] is raised otherwise. This works fine, as long as the used [f] is implemented in OCaml. But when [f] is just an OCaml stub for a C-function, then the handler for SIG_ALRM is called (and hence the exception Timeout is raised) but no exception [Timeout] arrives anywhere and hence the code of [f] runs as long as it needs ignoring any timeout.

My theory is that when SIG_ALRM is raised during execution of C-code, than
the [raise Timeout] statement raises an exception but as there is no context set up for exception handling this exception just gets lost.

Is this theory correct and are there any suggestions how to work around this problem (implementing timed execution of arbitrary code).

thnx in advance

christian