I've been trying without success to hack the following problem for about a
week:
Compiling emacs 20.3 seems to build temacs fine, runs it, loads a bunch
of list code, dumps (unexecs) to produce an emacs binary. When the
makefile runs the emacs binary:
./emacs -q -batch -f list-load-path-shadows
the dumped state information seems to not be loaded. I see:
Bare impure Emacs (standard Lisp code not loaded)
and it sits there waiting for input. If I send EOF, I can finish the
build and install, but the installed emacs just displays the above
output in the minibuffer and isn't otherwise useful.
Emacs was configure using:
./configure --prefix=/pkg/gnu/emacs-20.3
and compiled using "make". CC, CFLAGS, LDFLAGS, LDLIBS and suchlike
were unset during both configuration and make.
I can work around this by aliasing emacs to "emacs -l loadup" but this
is far too slow to be actually useful.
My system:
Alpha Noname (166MHz, uniprocessor)
emacs 20.3
egcs 1.1b (aka egcs-2.91.57 19980901 release)
binutils 2.9.1.0.15
glibc-2.0.7pre7
linux kernel 2.1.123
ncurses 4.2
_NO_ X windows stuff is installed. (This is a headless machine.)
This is a fairly complete Linux system, and there is not one bit of
code on it (other than MILO) which I did not compile from scratch using
the above tools (including the tools themselves). So, I'm inclined to
think this is a "Doug can't compile emacs right" problem and not a
"Doug doesn't have a sane build environment" problem.
Any solutions, suggestions for experiments to try, comments, whacks
with the Cluebat or questions via email or this 'froup will be
warmly welcomed. In any case, thanks for reading this far.
(Also, I keep hearing people mention a glibc-2.0.7 release, and/or
a glibc-2.1. Anybody have a URL?)
--
Headers munged to avoid email spam. Mail to henke at phoenix dot net.