Re: PATCH to support native Windows builds

[Mark Mitchell <mark at codesourcery dot com>]

> For last 4 years,  I have been bootstrapping gcc and building binutils with a native
> (msvcrt.dll-dependent) gcc and binutils toolchain.  

So, you made different changes to collect2.c than we did, or avoided 
using that code altogether -- which isn't possible for some cross 
compilers.

DJ's objections to the patch I posted are principled, but in my opinion, 
misguided.  Not being willing to expose internal libiberty interfaces to 
support Windows out of fear that these interfaces would not then be 
supported on DOS and/or AmigaOS strikes me as cutting off your nose to 
spite your face.

That said, I'm utterly dispassionate about how we get this done.  If 
there are problems with our patches (although our customers haven't told 
us about them), that's fine.  If there's a better interface, that's fine 
too.  If DJ will accept Ian's patches, that's fine by me.

The broader point is that there are patches out there that get GCC to 
work on Windows, without Cygwin, and that lots of people want to do 
that.  However, as a community, we seem to have failed to have gotten 
any such patch into GCC.  That means that multiple parties are creating 
and maintaining different patches to add this support.  And I'm not keen 
on spending a lot of time on working out another interface and more 
implementations; we've already got ways of doing this that (perhaps 
modulo a bug or two) work.

Danny, would you post the full patch so that DJ can more easily consier 
it?  DJ, will you accept Ian's patches?

If not, I guess I'll plan to put either Ian's patches or ours somewhere 
else in gcc, since DJ has indicated that he doesn't mind if we bypass 
libiberty.  Even this ugly alternative seems better than having lots of 
variants on this functionality out there in the world.

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com