On 1/23/14 2:17 PM, Brad Anderson wrote:
On Thu, Jan 23, 2014 at 2:55 PM, Andrew Edwards <[email protected] <mailto:[email protected]>> wrote:On 1/23/14, 2:01 PM, Walter Bright wrote: I agree, I don't know what's wrong with what we had before: 1. All pull requests get merged to master 2. Create 2.065 branch 3. Cherry-pick from master to 2.065 as required 4. Tag 2.065.whatever as releases get done on that branch Easy, simple. All these other procedures seem like massive over-engineering to me. Good to go... I for one did not see either of you weigh in on the proposal when Brad Roberts Brad Anderson :P made it (http://forum.dlang.org/post/__CAFU1Uzpm4DBADOxMjcJ_Guj1=__T8BQ4nPb5OEbADNbUQDD2ijuQ@__mail.gmail.com <http://forum.dlang.org/post/CAFU1Uzpm4DBADOxMjcJ_Guj1=t8bq4npb5oebadnbuqdd2i...@mail.gmail.com>). I decided to use it because, compared to the alternative of trying to convince volunteers to do something they do not want to, it would be much simpler for me to follow this scheme. I wish I would have thought to email Brad directly (sorry, Brad) to make sure he saw it and could weigh in. Especially since apart from you he's really the only other person that needs to change anything to adopt this workflow. To me there is no difference between the two processes, except the "we've always done it this way syndrome". Fixes are generated from release tags into a hotfix branch. Once the fix is released, we merge it back into master, remove the branch and move on. I am preparing both releases and hotpicks so I don't see any extra work being generated for the devs. The only chance I see on your parts is the need to change the tester scripts to point search for and test "hotfix" and "release" branches if they exist. I'm not the person doing that so I might have an overly simplified view of your processes but I really don't see the big deal. If Brad Roberts decides it's too hard for whatever reason we should be able to just change the workflow over to use a versioned branch name and dropping the step where the branch is deleted. Then the hotfix process would just checkout the versioned branch (and skip the delete as well). I like the tag and delete method better but we can't sacrifice the autotester for that.
The problem is that as specified, _every_ fix requires also setting up builds in the auto-tester (regardless of who does it). That should be once per maintained version. Deleting and recreating is a waste of everyone's time.
It's not just me that's affected. Anyone who wants to test releases as they're being built has to carefully track what branch to use when, which is tedious and a waste of time.
Also, what if we decide to patch two past releases, does that happen serially, using the release branch name for each of the versions one at a time? Also stupid and a waste of time.
Should I continue or is it obvious now? _______________________________________________ dmd-beta mailing list [email protected] http://lists.puremagic.com/mailman/listinfo/dmd-beta
