xfs
[Top] [All Lists]

Re: xfsprogs: update version for 3.2.0-alpha1

To: Eric Sandeen <sandeen@xxxxxxxxxxx>
Subject: Re: xfsprogs: update version for 3.2.0-alpha1
From: Ben Myers <bpm@xxxxxxx>
Date: Mon, 23 Sep 2013 10:07:27 -0500
Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>, Dave Chinner <david@xxxxxxxxxxxxx>, rjohnston@xxxxxxx, xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <52404A6E.4070306@xxxxxxxxxxx>
References: <20130916205637.GD1935@xxxxxxx> <20130916223855.GF19103@dastard> <20130923122623.GA2199@xxxxxxxxxxxxx> <52404A6E.4070306@xxxxxxxxxxx>
User-agent: Mutt/1.5.20 (2009-06-14)
Hi Gents,

On Mon, Sep 23, 2013 at 09:04:30AM -0500, Eric Sandeen wrote:
> On 9/23/13 7:26 AM, Christoph Hellwig wrote:
> > On Tue, Sep 17, 2013 at 08:38:55AM +1000, Dave Chinner wrote:
> >> On Mon, Sep 16, 2013 at 03:56:37PM -0500, Ben Myers wrote:
> >>> xfsprogs: update version for 3.2.0-alpha1
> >>
> >> I'd say this is a major feature and infrastructure
> >> update across the entire xfsprogs package, and in that case a
> >> PKG_MAJOR bump is warranted, not PKG_MINOR.
> >>
> >> i.e. We're shooting for a 4.0 release, not 3.2...
> > 
> > I tend to disagree with the 4.0 bump.
> > 
> > 2.0 was when the new xattr ABI was introduced, and 3.0 was when we
> > pulled fsr over from xfsdump to xfsprogs as well as drastically reducing
> > the amount of installed headers.
> > 
> > While the v5 support is a major internal change I think 3.2 would fit
> > better for this.
> 
> *shrug* TBH I Don't care a whole lot.  Externally for old users in theory
> it shouldn't be a big change.  But internally it's huge, and it enables
> a new disk format, so ... well, I don't want to bikeshed it too much.
> 
> I'd mostly like to see _something_ w/ a version number on it so distros
> can easily start to pick it up in testing repos.

I have no strong preference... there are plenty of letters in the alphabet.
 
> > I'd also be tempted to just cut 3.2.0 instead of an alpha version - just
> > keep the v5 support experimental, maybe under a configure option.
> 
> But so many changes are already made throughout the codebase, I think firing
> off a point release with half-baked V5 support seems weird at this point.
> 
> IOWs, aside from the V5 work I'm not sure anything merits a point release.

I do tend to agree with Eric that it is a good idea to do an alpha release
though.  A configure option is an intersting idea too, but that's not how it's
coded today.  Right now it's just a very loud warning when you create a
filesystem with crc=1.  That's probably good enough.

How about we just do a 3.2 alpha now to get something out there, and if after
all the painting is finished and y'all still want a 4.0 bump, we'll do one.  ;)

The major constraint being... we don't want to go backward.

-Ben

<Prev in Thread] Current Thread [Next in Thread>