From: Eric Sandeen <sandeen@redhat.com>
Subject: Re: [PATCH, RFC] mke2fs: get device topology values from blkid
Date: Fri, 18 Sep 2009 11:57:04 -0500
Message-ID: <4AB3BBE0.90105@redhat.com>
References: <4AB2B6B9.7010506@redhat.com> <20090918055547.GO2537@webber.adilger.int> <4AB39375.6040506@redhat.com> <20090918142024.GU2537@webber.adilger.int> <20090918164310.GD26991@mit.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: Andreas Dilger <adilger@sun.com>,
	ext4 development <linux-ext4@vger.kernel.org>
To: Theodore Tso <tytso@mit.edu>
In-Reply-To: <20090918164310.GD26991@mit.edu>
Sender: linux-ext4-owner@vger.kernel.org

Theodore Tso wrote:
> On Fri, Sep 18, 2009 at 08:20:24AM -0600, Andreas Dilger wrote:
>> On Sep 18, 2009  09:04 -0500, Eric Sandeen wrote:
>>> Andreas Dilger wrote:
>>>>> +	rc = blkid_probe_set_device(pr, fd, 0, 0);
>>>> Is this in the e2fsprogs blkid code yet?  I'm guessing not, since you
>>> right, not yet.  I don't know what the long-term plan is for e2fsprogs  
>>> blkid; keeping 2 trees in sync seems like a lot of work w/o much gain...
>> I thought that was Ted's plan?  We usually recommend people to update to
>> the latest e2fsprogs, yet not everyone will be able to upgrade to the
>> latest util-linux, especially if it is pulling in other libraries that
>> can cause package conflicts (e.g. with xfstools or similar).
> 
> I wasn't planning on keeping the blkid in e2fsprogs in sync with
> util-linux-ng; however, I did want to allow people who wanted to
> recompile the latest version of e2fsprogs (to pick up critical bug
> fixes, for example) on existing distributions (you know, for those
> silly Enterprise Linux Distro customers running RHEL 5 or SLES 11 that
> pay some of our collective salaries :-) to be able to do so.
> 
> If there is an issue that not all versions of util-linux-ng have the
> topology calls available, we'll need some autoconf magic, of course
> --- especially if some community distro's have shipped or will have
> shipped versions of util-linux-ng w/o the topology functions.

Yes, I'll do that for my changes in the next patch I send (Fedora is in 
exactly this boat).

> I don't particularly much care about screwing over a Linux From
> Scratch or Gentoo user who refuses to upgrade.  I care a tiny bit more
> about supporting a community distro (it's nice if we can support a
> given Community distro for at least a 12 months or so before we start
> making life difficult for them).  But to make life easier for
> Enterprise Linux users, and the Level 3 help desk people at IBM,
> Novell, and Red Hat who might need the latest version of e2fsprogs to
> fix some customer's file system corruption problem, we should provide
> an easy way for them to download and update that customer to the
> latest version of e2fsprogs.  The simplest way to do that is to stub
> out the new blkid topology functions has null inline functions in
> blkid.h.

I guess I'd still like to see lib/blkid move to lib/local-blkid/blkid or 
something, so that when --disable-libblkid is passed, we will never find 
local headers.  Without the --disable-libblkid, we would add 
-I./lib/local-blkid/ to the include search path.  Ditto for libuuid.

That way you don't even have to keep up with the stubs, right?  And 
autoconf tests for new interfaces will just gracefully fail if building 
everything against the in-tree headers.

Or if my build-fu is weak and there's a better way, I'm all ears.  :)

-Eric

> Does that make sense to everyone?
> 
> 						- Ted