From: Jens Axboe <axboe@kernel.dk>
To: Daniel Gollub <d.gollub@telekom.de>, fio@vger.kernel.org
Subject: Re: fio headers as build dependency for external-ioengines
Date: Wed, 30 Apr 2014 08:29:15 -0600 [thread overview]
Message-ID: <536108BB.7030502@kernel.dk> (raw)
In-Reply-To: <1398849904-6013-1-git-send-email-d.gollub@telekom.de>
On 04/30/2014 03:25 AM, Daniel Gollub wrote:
> Hi,
>
> I am working right now in preparing for the Ceph project an external
> fio ioengine, which depends on Ceph-internal libraries. So I thought
> about building this ioengine as an external ioengine. (Thats why I
> send couple of weeks/month ago this C++-compiler-issues-fixes for
> fio.h)
>
> Since fio is not (yet) distributed with a -dev/-devel package in
> linux distribution and was very likely not yet consider doing so
> I wanted to start the discussion here if this in interested of the
> project "exporting"/"installing" the required fio headers to build
> external ioengines. So fio can be later a "packaged" build-dependency.
>
> For time-being I thought about solving this like this:
>
> One can introduce in the external project's build-env a parameter
> like: --with-fio=~/projects/fio/ and the resulting build-env will
> pass this path as include directory as compiler flags for the
> external ioengine build.
>
> Unfortunately there are some issues with double-declaration and
> one ambiguous header file name (e.g. time.h - see provided patch)
>
> Double-declaration issues of defines/macros/... I hit while building
> the Ceph filestore external ioengine:
>
> * ARRAY_SIZE
> * CONFIG_CPU_COUNT
> * le16_to_cpu
> * le32_to_cpu
> * le64_to_cpu
>
>
> If you want to reproduce the described issue with ceph,
> checkout following revision:
> https://github.com/gollub/ceph/commit/079c08c67f811c0173e70201dc91168b60d2e86b
> ( Working branch: https://github.com/gollub/ceph/tree/fio_filestore_v2 )
>
> And run:
> ./configure --with-fio-dir=/path/to/fio/
> cd src
> make libfio_ceph_filestore.la
>
>
> This could be solved/workaround on the external ioengine side, as well
> as in the fio header. One possibility on the fio side could be to relax
> fio.h header a bit, so only mandatory declarations and further
> includes are part of it to successfully build an external-ioengine.
> I am happy to help on this if fio decides to go this route.
>
> Thoughts?
It'd be a useful thing to be able to do. I applied your time patch,
looks fine to me. Are you going to be able to work through the other
issues? I think we just need to privatize some of the names a bit more.
--
Jens Axboe
prev parent reply other threads:[~2014-04-30 14:29 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-04-30 9:25 fio headers as build dependency for external-ioengines Daniel Gollub
2014-04-30 9:25 ` [PATCH] Rename time.h for third-party include of fio.h Daniel Gollub
2014-04-30 14:29 ` Jens Axboe [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=536108BB.7030502@kernel.dk \
--to=axboe@kernel.dk \
--cc=d.gollub@telekom.de \
--cc=fio@vger.kernel.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.