From: Gabriel Scherer <gabriel.scherer@gmail.com>
To: Jordan W <jordojw@gmail.com>
Cc: "caml-list@inria.fr" <caml-list@inria.fr>
Subject: Re: [Caml-list] Problem with native compilation of mlpacks in ocamlbuild.
Date: Wed, 5 Aug 2015 07:00:55 +0200 [thread overview]
Message-ID: <CAPFanBEKY4A-0byw_9ajjPszg4UYGtCrWxNU5YjeX05sopL70w@mail.gmail.com> (raw)
In-Reply-To: <CAPOA5_6NonZsx_p8zD-7YNMwcPgowkxfYLW2ZTZVzF_1Dbwwyw@mail.gmail.com>
The fact that native packing requires -for-pack is not a bug of
ocamlbuild. The tool mirrors a design choice of the compiler itself,
where native and bytecode packing behave differently for technical
reasons (native compiled objects cannot be portably renamed, so their
final-destination name must be given from the start).
I am able to reproduce the issue, but it fails in both bytecode and
native, and a workaround (that is not usable in any case) it to add
"path/to/depOne" to the list of included directories of the build, by
invoking with (ocamlbuild -I path/to/depthOne) or having
("path/to/depthOne": include) in _tags.
On Mon, Aug 3, 2015 at 7:38 AM, Jordan W <jordojw@gmail.com> wrote:
> It is well known that there are unresolved issues with native compilation of
> mlpacks.
>
> Consider an mlpack that successfully byte code compiles:
>
> path/to/depOne.mlpack consisting of:
>
> depOne/A
> depOne/B
>
> Where depOne/a.ml contains
> let foo = "hi"
> Where depOne/b.ml contains
> let foo = A.foo
>
> While this byte code compiles without issue, modules inside of mlpacks will
> not be *native* compiled with their respective -for-pack X. To correct this,
> it has been suggested that a _tags file be created with the following:
>
> <path/to/depOne/**/*.cmx>: for-pack(DepOne)
>
>
> With this, both native and byte compilation succeed for the previous
> example.
> However, *because* B references A, then if B is located in another directory
> within the depOne root, native compilation will once again fail, even though
> byte compilation succeeds - even with all of the special hacks that have
> been suggested (_tags etc).
>
> For example:
> Consider if path/to/depOne.mlpack consisted of the following items, pointing
> to the new respective locations of A, B where B still refers to A as it did
> before.
>
> depOne/A
> depOne/deeper/B
>
> In this case, native compilation fails, and byte compilation succeeds.
>
> The errors that I see are:
>
> File "path/to/depOne.cmx", line 1:
> Error: Files path/to/depOne/deeperDir/b.cmx
> and path/to/depOne/a.cmx
> make inconsistent assumptions over implementation A
> Command exited with code 1.
>
>
> Can anyone suggest a fix for this?
>
> Thank you,
> Jordan
prev parent reply other threads:[~2015-08-05 5:01 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-08-03 5:38 Jordan W
2015-08-03 12:57 ` Ashish Agarwal
2015-08-04 20:39 ` Jordan W
2015-08-04 21:19 ` Ashish Agarwal
2015-08-05 5:00 ` Gabriel Scherer [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=CAPFanBEKY4A-0byw_9ajjPszg4UYGtCrWxNU5YjeX05sopL70w@mail.gmail.com \
--to=gabriel.scherer@gmail.com \
--cc=caml-list@inria.fr \
--cc=jordojw@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