Troubles compiling GNU Binutils
Fri Feb 13 01:06:00 GMT 1998
> If you get this to work, I'd love to hear about it. I called and spoke
> with someone at Cygnus about support, since I couldn't get the build to
> work to save my life. I learned two things in my conversation with
> someone at Cygnus: (1) they aren't interested in supporting anything but
> large development shops; and (2) you can't build the gcc toolchain using
> the Cygwin32 tools.
It is possible...you just need a lot of bleeding edge tools.
I have recently build a m68k-coff compiler c/w stdc++ and newlib using:
Gnu-win32 beta 18 tools (the basic building blocks)
Sergey Okhapkin's Coolview cygwin.dll replacement
EGCS-1.0 binaries from Mumit Khan
a build and install of "bison-1.25" which seems to be needed for doing
EGCS (version obtained via CVS at around 30/1/98)
binutils-184.108.40.206.18 obtained a while back from
ftp.yggdrasil.com/private/hjl (see the EGCS FAQ for details)
newlib-1.8.0(with Joel Sherril's RTEMS 4 patches)(which I think fix a few
build problems but I'm not sure)
There's a few hacks required along the way to get EGCS to build GCC/EGCS
properly(thanks to Mumit for helping with that one!) and the makefile needs
to be tweeked in a few places after configure....but it can be build
Alternatively you could pick up my "gnu-win32 source" build aimed primarily
at the ColdFire from http://www.calm.hw.ac.uk/davidf/coldfire.htm or wait
around a week until I clean up my build/install tree and release the new
version which handles non coldfire targets neatly.
> I've switched to a FreeBSD box and have had no difficulty getting the
> toolchain to build. Make sure you install gnu make, since the berkeley
> make that comes with FreeBSD won't correctly build the toolchain.
hmmm. I agree with you there...I test all "development" versions of EGCs on
Linux first to check that they build and compile code correctly before I
even think about win32...why? Complete toolchain rebuild takes 1hour 1 mins
under Linux on my P100 64MB SDRAM or 3-4hours on an AMD K6-200 under
NT....The resulting toolchain is just as good and almost as quick but any
errors in the build process are expensive!
David Fiddes, CALM Software Production Officer
Department of Mathematics, Heriot-Watt University, Edinburgh
email D.J.Fiddes@hw.ac.uk - Tel: +44 131-451-3251
P.S. I thought that the RTEMS list was firstname.lastname@example.org these days? My
mailer has dumped this thread in the wrong folder...grumble grumble
More information about the crossgcc