Bjarni Ingi Gislason wrote: > 2) with the commit, the config.log contains > > GL_GENERATE_ERROR_H_FALSE='' > GL_GENERATE_ERROR_H_TRUE='#' > > and the Makefile contains > > lib/error.h: $(top_builddir)/config.status > rm -f $@
That's because you happen to do this build on a platform which already has /usr/include/error.h. On the other platforms error.h gets created in the build directory, from error.in.h. You cannot have a reliable build system if the same include file basename (here: "error.h") is used to denote two different .h files, one that declares error_at_line, and one which declares groff C++ classes. This will *never* be reliable, regardless how to attempt to arrange the -I options in the AM_CPPFLAGS. Bruno
