martillo@mit-athena.ARPA (Joaquim Martillo) (12/06/84)
A few weeks ago, a fix to get CCAEmacs to load in the .elisp_init file
was posted to net.emacs. The fix only worked if the .elisp_init file
was linked to .elisp_init.l. If the function stream_given_loadstring is
altered so that the code looks something like this, emacs will load in
the .elisp_init file without a link to .elisp_init.l. In fact, if the
full path name to a file is fiven, any elisp file could be loaded even
if its name does not terminate in .l.
if (is_full_pathname(name))
{
searchlist = ads_cons_protect(Null_String,Nil_Symbol);
Pop_Elisp_Val;
goto no_extension;
}
else {
searchlist = ads_cons_protect(Null_String,searchlist);
Pop_Elisp_Val;
}
/* find out whether the filename already has the proper extension */
/* attached. If so we set extension to "" */
acq_sinfo(name,&schars,&slen);
acq_sinfo(extension,&echars,&elen);
if (elen >= slen) {
goto ok_extension;
}
while (elen > 0) {
if (echars[--elen] != schars[--slen]) {
goto ok_extension;
}
}
#if DEBUG
fprintf(dbgfp,"Setting extension to blank\n");
#endif DEBUG
no_extension:
extension = Null_String;
ok_extension :martillo@mit-athena.ARPA (Joaquim Martillo) (12/06/84)
stream_given_loadstring is in bio.c.
massar@godot.UUCP (J.P. Massar) (12/07/84)
In article <59@mit-athena.ARPA> version B 2.10.2 9/17/84; site godot.UUCP version B 2.10.1 6/24/83; site mit-athena.ARPA godot!mit-eddie!genrad!decvax!mit-athena!martillo martillo@mit-athena.ARPA (Joaquim Martillo) writes: >A few weeks ago, a fix to get CCAEmacs to load in the .elisp_init file >was posted to net.emacs. The fix only worked if the .elisp_init file >was linked to .elisp_init.l. If the function stream_given_loadstring is >altered so that the code looks something like this, emacs will load in >the .elisp_init file without a link to .elisp_init.l. In fact, if the >full path name to a file is fiven, any elisp file could be loaded even >if its name does not terminate in .l. > > > if (is_full_pathname(name)) > { > searchlist = ads_cons_protect(Null_String,Nil_Symbol); > Pop_Elisp_Val; > goto no_extension; > } > else { > searchlist = ads_cons_protect(Null_String,searchlist); > Pop_Elisp_Val; > } > > > /* find out whether the filename already has the proper extension */ > /* attached. If so we set extension to "" */ > > acq_sinfo(name,&schars,&slen); > acq_sinfo(extension,&echars,&elen); > > if (elen >= slen) { > goto ok_extension; > } > while (elen > 0) { > if (echars[--elen] != schars[--slen]) { > goto ok_extension; > } > } >#if DEBUG > fprintf(dbgfp,"Setting extension to blank\n"); >#endif DEBUG >no_extension: > extension = Null_String; > >ok_extension : I am the author of Elisp. This fix renders invalid the documentation in the Elisp manual, page 1-25. (Section 1.23.4 Loading Files) As I see it, the problem is not that one wants to be able to load .elisp_init, it is that the Emacs documentation is wrong in that the file should be called .elisp_init.l . By causing different behavior when a full pathname is given versus when a filename is given (where Elisp assumes the current working directory) leads to confusion in my mind. The current way to load an arbitrary file (which you generally don't want to do a lot anyway) is: (let ((temp *lisp-extension*)) (unwind-protect (progn (setq *lisp-extension* "") (load <myfile>)) (setq *lisp-extension* temp))) which could easily be subroutine-ized. Another algorithm might be to have Elisp look for <myfile> if <myfile.l> does not exist. But should it first scan all the directories in *LOAD-PATHNAMES* for <myfile.l>? Caveat emptor: There is no guarentee that this or any other bug fix / documentation fix will be incorporated into the next CCA Emacs release. JP Massar ihnp4!godot!massar massar@cca-unix
martillo@mit-athena.ARPA (Joaquim Martillo) (12/09/84)
Another possibility for fixing the problem and still allowing extension for full pathname files would be to do an rindex to the last / in the path and then do a strcmp whether the filename was .elisp_init. If the last path-element is .elisp_init, then the goto to no_extension could be executed. Then pathname extension works for all full path name files except .elisp_init. The argument sounds theological, no other emacs . file has an extension (eg .emacs_vars, .emacs_init .emacs_keys). If the documentation is wrong, then emacs should be changed because emacs only tries to load .elisp_init and not .elisp_init.l. Perhaps Zimmerman would like to comment.
martillo@mit-athena.ARPA (Joaquim Martillo) (12/09/84)
The necessary code so that the full path-name files can be properly
extended and the .elisp_init file can be loaded without the extension is
if(!strcmp(rindex(Get_string_chars(name, '/')), "/.elisp_init"))
{
goto no_extension;
}
.
This would substitute for the unconditional goto in my original fix.
Without either flavor of goto no_extension, but with the previously posted
load fix emacs would try to load .elisp_init and elisp would load
.elisp_init.l. I think there would be no need without the changes I
suggested to have a .elisp_init file. Actually, I prefer Massars's
documentation change. CCAEmacs has too many strcmp's as it is.