So you'd build 2 sets of test rpms this week:
-0.0.9.1
-0.0.9.2
+1.6.8.91
+1.6.8.92
-we decide it's fine then and we release
+we decide it's fine then and we release:
-0.1.0
+1.6.9
+
+If there are critical hotfixes that need to be released to customers
+(e.g. data corruption fixes) then point releases would be created:
+
+1.6.9.{1,2,3,...}
We go on developing with
-0.1.0.{1,2,3,4,...}
+1.6.9.{91,92,93,94,...}
as test releases and then we release:
-0.1.1
+1.6.10
-The 0.1 sequence is an unstable sequence, like 2.5 for the kernel is.
-So we expect lots of 0.1.X releases leading up to a stable 0.2 (or
-1.0) at the time of deployment.
+The 1.7 sequence would be an unstable sequence, like 2.5 for the kernel
+is. So we expect lots of 1.7.X releases leading up to a stable 1.8 (or
+2.0) at the time of deployment.
CVS
===
major.minor.patch.test
Such versions will be tagged in CVS as:
- v1_2_11_7
+ v1_6_9_97
and referred to as:
- 1.2.11.7
+ 1.6.9.97
encoded as:
- 0x01021107
+ 0x01060961
Usage:
------
- odd : when a new development cycle starts after a release
3. patch:
- when a development snapshot or release update becomes available
- - all these are announced on lustre-devel@lists.sf.net
+ - all these are announced on lustre-{announce,devel}@clusterfs.com
4. test:
- when developers feel it is time to exchange a named version
2. For three digit releases/tags the code should perform
according to the announcement.
-Moving tags
------------
-
-The last stable release will be tagged: CVS tag "t_last_stable"
-The last operational development snapshot will be CVS tag "dstable"
-
Branches
--------