[Opensim-dev] to branch or to break?

[Stefan Andersson]

> > as much as we all want osgrid to move forward, we should really tag> > 2060 as osgrid-stable-grid-0.4 and 2204 as osgrid-region stable-0.4> > and tell all people on osgrid to use THAT instead of trunk until we> > figure out what the hell is wrong.> > Though, we won't figure out what is wrong if everyone hangs back at safe> versions.
Well, our need to have our debuggers using the latest rev should be balanced against our need to push forward. I think this will mean we all try to make sure it moves forward.
 > > Also, if we want to supply some kind of stable experience, we should> > start _branching_ as well, bugfixing for that branch and selectively> > merge the bugfix back into trunk. You want stability or latest> > features? You can't have both.> > I really think history has shown that merging + svn = more pain and> agony than anyone really can put up with.
Well, these _branches_ aren't meant to be merged back, that's what I meant with 'selectively merge the bugfix' - either you do an actual diff merge, or you just re-fix the bug in trunk.
 
My point is that that 'stable branch' should only be _bugfixed_. New features goes into trunk.
> > The situation we have now with the core devs (newbs and oldes)> > tip-toeing around, being afraid to introduce bugs, because anybody> > could be expected to pull trunk and run it is just absurd. Also, it> > leads to n00bs like myself not being able to learn by 'trial and> > error' because we don't tolerate error. And yes, I do mean 'trial by> > error' by committing - bughunting should be a community effort and> > thus a community learning experience.> > Keeping the quality of trunk high is more important on the development> front than the user front. We've got 13 committers now, all working on> different schedules. If someone breaks trunk in a bad way, the other 12> can't get things done. Mistakes happen, but testing prior to commit> really needs to become part of our mantra if we're going to continue to> work as productively with so many developers. :)
 
As much as I do agree, testing now comprises so many scenarios that just testing for all scenarios would take longer time than actually coding, especially when working with functionality that touches grid code.
 
This project is alpha. We are blessed with tech-savvy people that have set up grids (several, in different environments and with different degree of heterogenity) - I say we _use_ them as a development resource, instead of devoting precious dev time to manual setup and testing in _fear_ of them. It should be up to each grid and region maintainer to test revisions and decide whether to move forward; what the devs can do is to provide _information_ about different revs, what to stay clear of, what's considered safe, and what needs testing.
 
/Stefan
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-dev/attachments/20071102/7dec4891/attachment-0001.html>