On Fri, 6 Aug 2010, kamaraju kusumanchi wrote: > > Kamaraju, > > > > Overall I like your plan. And I'd like to help. > > > > I do not like starting with version 3.0.6. I think such an old version > > is unlikely to attract many users and hence testing will be suboptimal. > > In addition, upstream reports that upgrading to 4.5 is currently broken > > (http://sagemath.org/mirror/src/changelogs/sage-4.5.1.txt), so we > > know that older releases will incur substantial development challenges > > that even upstream is not supporting. Moreover, upstream releases very > > frequently (lately releases have occurred more often than once per month), > > so by the end of the squeeze+1 cycle we will experience many, many upgrade > > tests. So adding to the testing burden by doing a "dry run" with legacy > > versions seems to me to be a very inefficient use of volunteer time. > > Indeed, until the packaging process becomes very efficient (which might > > take substantial time), I think it would be smarter to conserve limited > > volunteer resources by not packaging some of the upstream releases. > > > > You have a point. But the way I see it is this. > > Sagemath is constantly updated at a rate greater than debian can cope > up. I highly doubt we will ever be releasing the .deb packages as fast > as they release the .tgz files. So, at some point we have to skip > releases and provide as latest debs as possible. I understand that. > > But now the situation is a bit different. Are we sure that we have all > the deps of sagemath packaged into Debian? If the answer is yes, then > I am happy to start with 4.5 right away.
I think there are a couple new dependencies that are not in Debian; there weren't any as of version 4.0 or so. I would recommend first getting sagemath working building the copies contained in the sagemath tarball, and then package them separately for Debian and switch over later in development (this is how I did the original development, and it was much easier to debug problems incrementally). I suspect that starting by doing the work incrementally with 3.0.6 first might be easier than starting with 4.5 to begin with. There's a good chance you'll want to switch tacts once you get the hang of it, but I think if you try migrating the current package to 4.5, you'll end up feeling overwhelmed by the problems and give up. Some partial progress of mine on updating direct to 3.4.1 (shortly before 4.0) is available, in case you find it useful (I don't think I was very far along): http://web.mit.edu/sage/www/sage-3.4.1-debian.tar.gz My experience is that one spends most of your time working on sagemath packaging on (1) debugging and (2) waiting for it to build (it took about 30 minutes to build on the server I was using). When I tried to update direct from 3.0.5 to 3.4.1, I found debugging problems resulting from upstream changes took most of the time. I bet it would be much easier when you can find the upstream change that caused the problem; since each sagemath version has relatively small changes, that can make life easier, especially if you're still getting used to dealing with the Sage build system. One thing that I should warn you about is that now Debian has substantially newer versions of various packages than Sage 3.0.5 was designed for, and in some cases that will break things. The current Sage 3.0.5 package was prepared for Lenny, and then tweaked a bit to keep it compiling on newer stuff. So it's possible that the incremental approach will prove to be painful and you don't want to do it. But if I were you, I would probably start by just trying to do 3.0.5 -> 3.0.6, just because I think that'll help build confidence and give you a better sense of the nature of the challenge than going straight to 4.5. But it's really up to you. I don't have the time to help more than just providing background information on how I did it and what problems I encountered. -Tim Abbott > If the answer is no then the next question is what is the minimal > version that we can package given the current set of packages > available in Debian. There is no clear cut approach. we need to go > back and forth a bit. We may need to file some ITPs and work on some > transitions which is where the team becomes important. > > As for the support requests from users, sooner or later they realize > that if there is a problem they have to go with the later version > anyway. A bit of that frustration is probably good as it will drive > some to come and take part in packaging sage for Debian. -- 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
