On Sun, 2006-06-25 at 11:08 +0200, Erwin Rol wrote: > On Sun, 2006-06-25 at 10:59 +0200, Ralf Corsepius wrote: > > On Sun, 2006-06-25 at 10:43 +0200, Erwin Rol wrote: > > > On Sun, 2006-06-25 at 07:41 +0200, Ralf Corsepius wrote: > > > > On Sun, 2006-06-25 at 04:55 +0200, Ralf Corsepius wrote: > > > > > > Ralf, > > > > > > > FYI: I've uploaded my FC5->Mingw cross toolchain to Packman. > > > > ftp://ftp.uni-erlangen.de/pub/mirrors/packman/fedora/5 > > > > > > I took a short look at your spec files, do I understand correctly that > > > you use the binary versions of mingw32-runtime and w32api to bootstrap > > > your compiler build ? > > Yes, I am using the original target libraries and build a sys-rooted > > cross-toolchains from it. > > > > That's basically the same approach, you'd apply to building a native > > GCC/binutils for hosts with closed source libc (e.g. commercial *nixes). > > What i did was rebuild the mingw32 and w32api from the source, and was > wondering how you solved the circular dependencies, The standard approach would be to build a minimal gcc first, use this gcc to build a libc, then rebuild a "full gcc". For newlib-based targets (such as rtems), a special exception exists: For them, newlib/libc can be built simultaneously with GCC. > of course when using > the binaries there is no circular dependency :-) Exactly. It also has another advantage: Nothing can be "more original" than the "original" - I.e. why rebuilding target libs when you can use the original files? Ralf -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
