On Thu, Mar 15, 2012 at 10:56 AM, Keshav Kini <keshav.k...@gmail.com> wrote: > Robert Bradshaw <rober...@math.washington.edu> writes: >> On Thu, Mar 15, 2012 at 3:34 AM, Dima Pasechnik <dimp...@gmail.com> wrote: >>> On Thursday, 15 March 2012 17:55:31 UTC+8, Jeroen Demeyer wrote: >>>> >>>> Okay, I'll think about your suggestion and changing the merger >>>> procedure. But I'll be honest that this is not too high on my list of >>>> priorities. >>> >>> >>> well, I think that Keshav's approach is very important if we want to >>> decrease the huge rate of bitrot we have now >>> with patches that did not make it into a release quickly. >>> As the current scheme of things destroys the history of development, it gets >>> hard to recreate the state of source when >>> the now bit-rotten patch has been working. >> >> +1. Note that this is rather orthogonal from the git vs. hg >> discussion, right now we're using hg as a glorified diff and patch >> (and periodically-constructed changelog). > > Just to be clear: the git vs. hg discussion, the destroying history of > dev releases discussion, and the patches vs. push/pull discussion are > all orthogonal to each other, though push/pull exposes the problems with > destroying history more clearly than the patches workflow does (as I > said in my reply to Dima). > > Of these three things, IMO the most important is patches vs. push/pull . > I really just made this thread in the hopes of getting the destroying > history problem out of the way so that it won't stand as a barrier to > switching to push/pull (though it is bad in its own right anyway, even > when using patches).
I was trying to avoid adding a fourth thing to the mix, but I think it's fairly relevant. Sage itself isn't under revision control, rather it's a collection of a bunch of binary blobs (which each have their own repositories). Now the largest one, of course, is the Sage library itself, but it's not entirely decoupled from the rest. This means that one can't "push/pull" Sage, and trying to do so with just the sage library without all the relevant spkgs is dangerous. Furthermore, the idea of basing 5.0.alpha0 off of 4.8.beta1, doing work on both, then "merging" is much more involved (separately merging the various repositories and manually updating the spkg blobs--much more than an hg/git merge + resolve conflicts). This hurts us when doing development as well, but it's usually not as bad because most changes are local to a single repository, or at most two. - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
