[erlang-bugs] Build fix for some architectures: x86_64 e.g. [resent]

Mikael Pettersson <>
Fri Dec 14 13:45:08 CET 2007


Christian Faulhammer writes:
 > Mikael Pettersson <>:
 > 
 > > Christian Faulhammer writes:
 > >  > Mikael Pettersson <>:
 > >  > > This is ambiguous. Please generate a standard patch
 > >  > > that applies with 'patch -p1'.
 > >  >  Better?
 > > Better.
 > > Now, could you please explain why this is needed?
 > > I.e., why does gentoo need this change while (apparently)
 > > everybody else Unix-like doesn't.
 > 
 >  To quote from the original bug report (in 2005, I was not around then):
 > "no, that's a crappy 'fix' as is the current amd64 solution
 > 
 > the reason for this is the build system doesnt setup TARGET properly
 > 
 > configure.in sets TARGET to $host which comes from --host=... and this
 > value is recorded in the toplevel Makefile ...
 > 
 > when the generate target is run in the toplevel Makefile, it drops all
 > settings and changes to erts/emulator which sources make/run_make.mk
 > which sources make/target.mk which goes 'TARGET is not set, lets guess
 > at it using config.guess' which then produces the garbage value
 > 'sparc64-unknown-linux-gnu' and 'x86_64-unknown-linux-gnu'"

The "garbage" being that you lose your -gentoo- marker?
Or do you need to force TARGET to work around the issue of
inconsistent config.{guess,sub} files in the source tree?
(see Sergei Golovan's message today)

Either way I now see the benefit of this change.

/Mikael



More information about the erlang-bugs mailing list