MinGW Toolchain

[Ralf Corsepius]

On Tue, 2007-11-20 at 10:29 -0600, Scott Newell wrote:
> At 10:09 AM 11/20/2007 , Ralf Corsepius wrote:
> >
> >On Tue, 2007-11-20 at 16:54 +0100, Wolfram Wadepohl wrote:
> >
> >> also cygwin was not as stable it should be.
> >... I would not subscribe to this.
> 
> I take it the old build_alias problem is now fixed?
Don't know. It's been years since I heard about the build_alias issue
for the last time.

All I can say:

- Wrt. building *nix->cygwin/mingw cross-toolchains, building cygwin
toolchains is one magnitude easier. The basic procedures are the same,
but the details on MinGW are nastier.

- Wrt. maintaining cygwin/mingw->rtems toolchains, cygwin is one
magnitude easier, because Cygwin has a centralized "distro" upstream. 
Problem with Cygwin is it's stagnating upstream. 
The converse applies to MinGW - It doesn't seem to have reasonable and
stable "distro" upstream. You end up building everything yourselves.

- Using cygwin is one magnitude easier, because it's more "*nix'ish"
than MinGW. The situation of MinGW is gradually improving, nevertheless
there still are many pitfalls.

- Cygwin is sharing some amount of sources with rtems (newlib). 
I.e. RTEMS buys in some amount of code and QA from Cygwin. This doesn't
apply to MinGW.

> Has anyone compared build times under MinGW to Cygwin?  Cygwin sure seems
> slow...
I am using neither. I am occasionally building *nix->MinGW/Cygwin
toolchains and am working on Canadian-cross building packages on Linux.


Have you tried to build MinGW, GCC, binutils, gdb under MinGW?
AFAIK, this isn't possible, but I could be wrong.

Ralf