David Holland <dholland-pkgtech%netbsd.org@localhost> writes: > On Fri, Oct 16, 2015 at 11:02:00AM +0200, Thomas Klausner wrote: > > How does this sound: > > > > * we define, per operating system/version, a default c++ compiler, > > let's say gcc-4.9 (or whatever we think is the best version on that > > platform). this version is pulled in whenever c++ is defined in > > USE_LANGUAGES. It should support c++11. > > > > * the user can override this default globally, before compiling packages > > I feel like in the long run we want/need to have support for settings > or values that pertain to whole disjoint groups of packages, both for > this and for certain similar builtin package issues. That sounds appealing, but I am wondering how it would work. Would there be some analysis on the graph of all possible packages and then there could be a different C++ compiler for each disconnected component? What kinds of problems would it solve (besides the A needs only X and B needs only Y), how often do we have those problems, and do you think it's worth the pain?
Attachment:
signature.asc
Description: PGP signature