caml-list - the Caml user's mailing list
 help / color / mirror / Atom feed
From: Goswin von Brederlow <goswin-v-b@web.de>
To: Florent Monnier <monnier.florent@gmail.com>
Cc: caml-list@yquem.inria.fr, Goswin von Brederlow <goswin-v-b@web.de>
Subject: Re: [Caml-list] How to pass C pointers to Caml
Date: Wed, 03 Mar 2010 12:02:46 +0100	[thread overview]
Message-ID: <87fx4hllbd.fsf@frosties.localdomain> (raw)
In-Reply-To: <201003021234.48943.monnier.florent@gmail.com> (Florent Monnier's message of "Tue, 2 Mar 2010 12:34:48 +0100")

Florent Monnier <monnier.florent@gmail.com> writes:

> Le mardi 2 mars 2010 11:19:58, vous avez écrit :
>> Florent Monnier <monnier.florent@gmail.com> writes:
>> > Le lundi 1 mars 2010 14:24:45, Goswin von Brederlow a écrit :
>> >> Florent Monnier <monnier.florent@gmail.com> writes:
>> >> > Le lundi 1 mars 2010 04:55:00, Jianzhou Zhao a écrit :
>> >> >> I have been calling OCaml code from C in my project.
>> >> >> The C code has some pointers to C structures.
>> >> >> I got 'seg fault' when calling the OCaml function receiving
>> >> >> C structure pointers.
>> >> >>
>> >> >> 18.7 at http://caml.inria.fr/pub/docs/manual-ocaml/manual032.html
>> >> >> gives the examples that pass int into OCaml. These examples work for
>> >> >> me. But, Does OCaml support to pass C structure pointers to OCaml?
>> >> >
>> >> > Yes it does. Just cast your pointer to the type value.
>> >> >
>> >> > In this tutorial there is an example "Pointers to C structures":
>> >> > http://www.linux-nantes.org/~fmonnier/OCaml/ocaml-wrapping-c.php#ref_p
>> >> >tr
>> >> >
>> >> > the pointer to a C struct is wrapped on the ocaml side by an abstract
>> >> > type called "t" here, and it is provided back to C with print_t /
>> >> > dump_ptr.
>> >>
>> >> The problem with this trivial approach is that ocaml can store the
>> >> pointer somewhere. When the C pointer is freeed then ocaml has a
>> >> dangling pointer. Worse, if the GC allocates a new heap then the pointer
>> >> might suddenly point into the heap and then BOOM.
>> >
>> > A lot of bindings wrap C pointer, it is known to be a technic that does
>> > work. Dangerous that's true, be if you are very careful, it works.
>> > What you can do is set the pointer to NULL when the struct is freed, and
>> > then each function that uses this struct pointer can first check if the
>> > pointer is NULL or not before to use it, and if it's NULL raise an
>> > exception.
>> 
>> let x = ref None
>> 
>> let called_function c_ptr = x := Some c_ptr
>> 
>> How will you get x to be Some NULL?
>
> I mean often in a C library pointers to C struct are just pointers to 
> something abstract because a lot of lib C API do hide the struct, and provide 
> a function to free the things pointed. So in your wrapper after you call the 
> destroy function you can set the pointer to NULL and in the other functions 
> test if the pointer is NULL before to use it to prevent a user of the wrapper 
> to call a function after having called the destroy function.
>
>>                                     Your C code does not know about the
>> copy. You need to wrap the C pointer into a custom or abstract block
>> first to be able to NULL it. A finalizer in a custom block can also be
>> helpfull here and free the pointer when ocaml no longer needs it.
>> 
>> >> It is better to put the pointer into an abstract or custom block.
>> >
>> > You can do this too.
>> 
>> Imho you must. Anything else is too dangerous.
>
> Yes it is dangerous, but as I explained with some C libraries this is the only 
> possible solution. If you can do it the other way, then do, while it's indeed 
> safer.

It is always possible to wrap the C pointer into an abstract or custom
block. The pointer can still become dangling when the C side frees the
target before ocaml forgets the block but, as the GC does not look into
abstract/custom blocks, it will never result in the GC exploding because
a dangling pointer points into the GC heap.

MfG
        Goswin


      reply	other threads:[~2010-03-03 11:02 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-03-01  3:55 Jianzhou Zhao
2010-03-01  6:22 ` [Caml-list] " Florent Monnier
2010-03-01 13:23   ` Jianzhou Zhao
2010-03-01 13:24   ` Goswin von Brederlow
2010-03-01 23:04     ` Florent Monnier
2010-03-02 10:19       ` Goswin von Brederlow
2010-03-02 11:34         ` Florent Monnier
2010-03-03 11:02           ` Goswin von Brederlow [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87fx4hllbd.fsf@frosties.localdomain \
    --to=goswin-v-b@web.de \
    --cc=caml-list@yquem.inria.fr \
    --cc=monnier.florent@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).