svn commit: r193241 - in head: . sys/sys

David Schultz das at FreeBSD.ORG
Thu Jun 4 16:28:30 UTC 2009


On Thu, Jun 04, 2009, M. Warner Losh wrote:
> In message: <20090604151959.GA26524 at zim.MIT.EDU>
>             David Schultz <das at FreeBSD.org> writes:
> : On Mon, Jun 01, 2009, Ken Smith wrote:
> : > It was noted we're close to running out of numbers we can use before we
> : > hit code freeze and the branch for the release.  Since we're entering
> : > code slush at the end of today in theory all changes that would warrant
> : > a bump in __FreeBSD_version are supposed to be done.  But it wouldn't
> : > surprise me if we have one or two or so things that come along between
> : > now and when we hit code freeze and the branch.  So we need to be a bit
> : > conservative with this.  Please be sure to coordinate anything that
> : > might require a bump in __FreeBSD_version with re@ from now on.  If it
> : > turns out things do come along that require bumps we'll need to "batch
> : > them up" having one bump represent several changes.
> : 
> : To avoid this sort of problem in the future, how about adding a
> : digit to __FreeBSD_version in 9-CURRENT?  Admittedly, a lot of the
> : bumps in 8.X were probably unnecessary, but it's good that people
> : are being cautious and documenting their incompatible changes.
> 
> We can avoid this problem by not being so bump-happy.
> 
> Adding an extra digit was painful when we did it before.  A number of
> subtle things broke (like the output of file).
> 
> Part of the problem here is that we want to ship FreeBSD 8.0 as
> '800100' which is just historical convention:
> 
>  * scheme is:  <major><two digit minor>Rxx
>  *		'R' is 0 if release branch or x.0-CURRENT before RELENG_*_0
>  *		is created, otherwise 1.
> 
> We could easily up that to '5' for the release so we have 499
> entries. There aren't so many things that depend on this convention in
> the tree (I couldn't find any in a quick, informal survey).

Sounds good.  We might want to use 7 or 8 instead of 5 to account
for the fact that changes requiring a bump are more frequent in
CURRENT than in STABLE.


More information about the svn-src-all mailing list