Mailing list for all users of the OCaml language and system.
 help / color / mirror / Atom feed
From: David Sheets <sheets@alum.mit.edu>
To: Malcolm Matalka <mmatalka@gmail.com>
Cc: O Caml <caml-list@inria.fr>
Subject: Re: [Caml-list] Including a C library statically in an Ocaml library
Date: Fri, 8 Jan 2016 15:24:52 +0000	[thread overview]
Message-ID: <CAAWM5Ty22Keg6vh_RqysUtA6sY7MYn1G5S2i1Boqg0TzVBTr7Q@mail.gmail.com> (raw)
In-Reply-To: <86egdsnlof.fsf@gmail.com>

On Fri, Jan 8, 2016 at 3:19 PM, Malcolm Matalka <mmatalka@gmail.com> wrote:
> David Sheets <sheets@alum.mit.edu> writes:
>
>> On Fri, Jan 8, 2016 at 2:14 PM, Malcolm Matalka <mmatalka@gmail.com> wrote:
>>> The core problem I am having is a C library I want to bind has a number
>>> of macros which I need the value of.  Here is how I am trying to solve
>>> it, but perhaps there is a better way:
>>>
>>> I have a small C library which gets compiled to libfoo.a which provides
>>> functions that return the macro values, like:
>>>
>>> int macro1() { return MACRO1; }
>>>
>>> I then have an ocaml library, called ofoo, that uses Ctypes to bind to macro1:
>>>
>>> let macro1 = foreign "macro" (void @-> returning int)
>>>
>>> Where I am having issues is this small library, I'd prefer it to not
>>> have to be installed on the system but just compiled into the Ocaml
>>> library so that a user just has to link against that library.  Right
>>> now, none of the symbols (macro1) are being included in the library, I'm
>>> guessing because the linker sees no direct use of them.  And I'm not
>>> even sure if I can get it included in the ocaml library.  I'm also not
>>> able to get the libfoo symbols linked into a final executable, I'm
>>> guessing for similar reasons.
>>>
>>> What are my options here?
>>
>> I'd recommend using ctypes 0.4+ stub generation support which can bind
>> macro values and detect struct layout at an early compilation stage.
>> You can see it in use to do this in my ocaml-unix-errno library
>> <https://github.com/dsheets/ocaml-unix-errno>. One benefit with this
>> approach is greater static checking and c <-> ocaml type safety.
>>
>> If you continue using dynamic bindings, two lonker flags may be of use to you:
>>
>> --no-as-needed : on recent Ubuntu distributions, gcc automatically
>> includes the --as-needed flag which drops symbols that are not
>> referenced. Unfortunately, clang does not understand this flag so you
>> need to have a conditional in your build system to detect the compiler
>> in use if you want a cross-platform library.
>>
>> -E : The Exports local symbols which are statically linked into a
>> binary into the dynamic symbol table so that they can be found with
>> dlsym.
>>
>> You can use these with gcc like -Wl,-E for example.
>
> Will these approaches require that I have the C library installed to
> compile against any binary using my library or will the symbols be part
> of the ocaml library?  In my current version I have a libfoo.a that gets
> created in the project and then linked against the library.

In the stub generation case, you will not have symbols that represent
the macros. Instead, C code will be generated which generates OCaml
code which contains the macro constants.

In the dlsym static linking case, you would build regular cm(x)a files
and a .a file and install them all with ocamlfind. Your users don't
have to know about C libraries or have them installed globally on the
system.

David

>>
>> Hope this helps,
>>
>> David Sheets
>>
>>
>>> If I've missed any useful information, let me know.  I haven't interoped
>>> much with C directory in Ocaml so I'm not sure what information is
>>> important.
>>>
>>> /Malcolm
>>>
>>> --
>>> Caml-list mailing list.  Subscription management and archives:
>>> https://sympa.inria.fr/sympa/arc/caml-list
>>> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
>>> Bug reports: http://caml.inria.fr/bin/caml-bugs

      reply	other threads:[~2016-01-08 15:24 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-01-08 14:14 Malcolm Matalka
2016-01-08 14:58 ` Christopher Zimmermann
2016-01-08 15:08   ` Malcolm Matalka
2016-01-08 15:15 ` David Sheets
2016-01-08 15:19   ` Malcolm Matalka
2016-01-08 15:24     ` David Sheets [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=CAAWM5Ty22Keg6vh_RqysUtA6sY7MYn1G5S2i1Boqg0TzVBTr7Q@mail.gmail.com \
    --to=sheets@alum.mit.edu \
    --cc=caml-list@inria.fr \
    --cc=mmatalka@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