> The FreeBSD core OS needs not much more than the basic language, > so miniperl's functionality comes vaguely close.
Solaris 8 has the same requirement: they have a very small subset of Perl, for Solaris "miniroot" setups, and for being able to run a few utilities that are really Perl wrappers for kernel interfaces. > Yes. FreeBSD needs to be able to cross-build. There are different kinds > of cross-builds, as Perl would see it; > > 1) Different CPU architectures > 2) OS upgrades (like FreeBSD-STABLE --> FreeBSD-CURRENT) > 3) Buildworlds after Perl upgrades (when miniperl has a > risk of failing). For 5.8.0 there is an (unfinished) push for better cross-compiles. I used that to successfully build a working "miniperl" for the ARM Ipaq PDAs while the Configure and the cross-gcc were on a x86 linux. Where the fun ended was building the extensions, including the Dynaloader, so I couldn't build a real "perl". See INSTALL/Cross-compilation. Another guy, Vadim Konovalov, is putting in capabilities for non-Configure cross-builds (for WinCE). For "extreme cross-building" there's also a makefile target called "microperl" which tries to stick to pure ANSI C. No need for Configure, the makefile is static and always available (Makefile.micro). It seems to work pretty well, topugh it has not been much tested, and it would definitely need a few bootstrapping rounds to be safe since it e.g. contain hardcoded constants for integer/pointer sizes, which means that in places like Alpha or "big-numbered" MIPSes theose sizes are severely wrong. Which reminds me: one complication that the build systems of various operations might not have given consideration but Perl has to: we really cannot (or, at least, should not) assume a particulart C compiler, such as the gcc. -- $jhi++; # http://www.iki.fi/jhi/ # There is this special biologist word we use for 'stable'. # It is 'dead'. -- Jack Cohen
