Le jeudi 01 mars, Michael Orlitzky a écrit: > On 03/01/2012 04:52 PM, William Stein wrote: > >> > >> I'm sorry, but I still fail to see how situation : > >> > >> sage contains all its deps (even if they're already there) > >> > >> will be any better for subtle bugs than situation : > >> > >> sage has the same deps, but doesn't manage them itself more > >> than by declaring them (and having the system packaging satisfy > >> them). > > > > Here's a rant about this (not written by me): > > > > http://wiki.sagemath.org/faq/bigsagerant > > > > This is wrong for two reasons. The first is that it claims, "one bad > attempt failed, therefore, success is impossible." > > The second is because it ignores the fact that you can specify > package versions in your dependencies (on Debian, and everywhere > else). If you specify that you need foo-x.y, your package manager > will install foo-x.y for you. It won't just use foo-w.z and break > everything. You can have both installed at the same time, even.
Let me add that you can also do the following : (1) specify your exact deps so they are satisfied upfront ; (2) end sage's build by tests, and declare the build a failure if the tests aren't 100% perfect. Take for example : https://buildd.debian.org/ There you see a real huge (really huge : several arch, tens of thousands of packages), and no package enters the distribution if it doesn't build properly (and if you consider point (2), that's with quite good confidence it will also give good results). Snark on #sagemath -- 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
