view src/pre-crt0.c @ 1687:2bee660c3046

* configure: Don't make the top-level Makefile read-only - people may want to edit the values of the path variables. Make path specification conform to GNU coding standards. * configure (long_usage): Remove all traces of old arguments from usage messages, and document the options we do accept in more detail: -with-x... and --srcdir. (options, boolean_opts): Deleted; we don't have enough options to make this worthwhile. (prefix, bindir, lisppath, datadir, libdir, lockdir): Deleted, along with the code which supported them; these should be set as arguments to the top-level make. (config_h_opts): Since this no longer doubles as a list of option names, make them upper case; this simplifies the code which uses them to build the sed command to edit src/config.h. Change the code which sets them. (cc, g, O): Don't allow the user to set these using options; they should be specified using `CC=' and `CFLAGS=' arguments to the top-level make. Just choose reasonable default values for them, and edit them into Makefile.in's default CC and CONFIG_CFLAGS values. (gnu_malloc, rel_alloc): Don't allow the user to set these using options; use them whenever the configuration files say they're possible. Simplify the argument processing loop. Don't accept abbreviations for option names; these might conflict with other configuration options in the future. Add some support for the `--srcdir' option. Check for the sources in . and .. if `--srcdir' is omitted. If the directories we will compile in don't exist yet, create them under the current directory. Note that the rest of the build process doesn't really support this. Edit only the top Makefile. That should edit the others. Edit into the makefile: `version', from lisp/version.el, `configname' and `srcdir' from the configuration arguments, `CC' and `CONFIG_CFLAGS' as guessed from the presence or absence of GCC in the user's path, and LOADLIBES as gleaned from the system description files. Simplify the report generated; it doesn't need to include any description of paths now. Make `config.status' exec configure instead of just calling it, so there's no harm in overwriting `config.status'. * Makefile.in (version, configname): New variables, used to choose the default values for datadir and libdir. Path variables rearranged into two clearer groups: - In the first group are the variables specified by the GNU coding standards (prefix, bindir, datadir, statedir, libdir, mandir, manext, infodir, and srcdir). - In the second are the variables actually used for Emacs's paths (lispdir, locallisppath, lisppath, buildlisppath, etcdir, lockdir, archlibdir), which depend on the first category. datadir and libdir default to directories under ${prefix}/lib/emacs instead of ${prefix}/emacs, by popular demand. etcdir and lispdir default to subdirectories of datadir. archlibdir defaults to libdir. The new installation tree is a bit deeper than it used to be, so use the new make-path program in lib-src to build them all. Always build a new src/paths.h.tmp and then move-if-change it to src/paths.h, to avoid unnecessary rebuilds while responding to the right changes. Remove all mention of arch-lib. Run utility commands from lib-src, and let the executables be copied into archlibdir when Emacs is installed. Add targets for src/Makefile, lib-src/Makefile, and oldXMenu/Makefile, editing the values of the path variables into them. Let lib-src do its own installation. (datadir): Default to putting data files under ${prefix}/lib/emacs/${version}, not /usr/local/emacs. (emacsdir): Variable deleted; it would only be confusing to use. (lispdir, etcdir): Default to ${datadir}/lisp. (mkdir): Use make-path for this. (lockdir): Do this in mkdir. (Makefile): New target. * configure (usage_message): Rename to long_usage.
author Jim Blandy <jimb@redhat.com>
date Sat, 12 Dec 1992 15:42:53 +0000
parents 3165b2697c78
children 695cf19ef79e
line wrap: on
line source

/* This file is loaded before crt0.o on machines where we do not
   remap part of the data space into text space in unexec.
   On these machines, there is no problem with standard crt0.o's
   that make environ an initialized variable.  However, we do
   need to make sure the label data_start exists anyway.  */

/* Create a label to appear at the beginning of data space.  */

int data_start = 0;