Re: [21.2.28] pdump doesn't work any more

mt> as it seems the behavior of --pdump has change. So now as opposed to mt> 21.2.26 I get temacs and xemacs both with the same size: mt> -rwx------ 1 thiessel thiessel 5159408 Feb 7 18:01 temacs mt> -rwx------ 1 thiessel thiessel 5159408 Feb 7 17:59 xemacs Yes, I did that... mt> The build process fails at mt> MAKE='make' EMACS='./src/xemacs' /bin/sh /ux/xemacs-21.2.28/lib-src/update-elc.sh mt> (using ./src/xemacs) mt> Recompiling in /ux/xemacs-21.2.28 mt> with /ux/xemacs-21.2.28/src/xemacs... mt> temacs can only be run in -batch mode. mt> *** Error exit code 255 I think you'll find if you do cd src; rm temacs xemacs; make you'll get a runnable xemacs. Well, we need to have an actual strategy for how temacs should differ from xemacs, or whether such a distinction is necessary. Several strategies are available: - xemacs/temacs looks at its executable name, and runs pdump_load only if not named temacs. This sort of thing is known to be evil. - we create files pre-dump.c and post-dump.c pre-dump.c: int load_dump_file_p = 0; post-dump.c: int load_dump_file_p = 1; We link temacs with pre-dump.o and xemacs with post-dump.o to create similar but different executables. This wastes 7mb of disk. - We use a command-line option -no-dumped-data-file, and change the Makefiles to do things like ./xemacs -no-dumped-data-file -batch -l ../lisp/update-elc.el Then we would never have a `temacs'. Olivier?