Re: Initial Paths Search

Adrian Aichner wrote:

>>>>>>"David" == David A Cobb <[email protected]> writes:
>>>>>>            
>>>>>>
>
>    David> Could you please advise me which lisp routines execute the startup
>    David> search for the load-path?
>
>See
>(defun paths-find-emacs-roots (root-p)
>and friends in
>lisp\setup-paths.el
>  
>
Well, actually in lisp\find-paths.el.  But I blundered my way around 
startup.el, find-paths.el, and UGH!! paths.el

Here's what I come up with.
;;; in startup.el, function 'normal-top-level . . .
    (setq default-directory (abbreviate-file-name default-directory))
    (initialize-xemacs-paths)                            ;;;; << This is 
just horrid [on a non-*nix system especially]

    (startup-set-invocation-environment)          ;;; here in 
startup.el, nothing useful

    (let ((debug-paths (or debug-paths
               (and (getenv "EMACSDEBUGPATHS")
                t))))

      (setq emacs-roots (paths-find-emacs-roots invocation-directory
                        invocation-name))
        ==> to find-paths.el.
        ;;; the only inputs used, other than the invocation-directory & 
-name are "configure-exec-prefix (== nil) and configure-prefix-directory 
(== "..\")

      (if debug-paths
      (princ (format "emacs-roots:\n%S\n" emacs-roots)
         'external-debugging-output))                                

      (if (null emacs-roots)
      (startup-find-roots-warning))                 ;;; << So, I'm 
doomed to a warning if I cannot get past this
      (startup-setup-paths emacs-roots           ;;; also here in startup.el
               user-init-directory
               inhibit-early-packages
               inhibit-site-lisp
               debug-paths)
      ====> to startup-setup-paths.  I'm excited to learn that it is 
idempotent -- I've never actually found a reason to use that word before!
For my purposes, the important thing is that it goes right to work on 
the package-paths.  After that, it calls . . .

        (setq lisp-directory (paths-find-lisp-directory roots)) ;;; Ah, 
that's in setup-paths.el
 calls    >>(paths-find-version-directory roots "lisp"  nil  
configure-lisp-directory)  ;;; Back to find-paths.el
 which calls >>> (paths-find-emacs-directory                  ;;; well, 
paths-find-emacs-directory knows how to use environment var's, but all 
it gets passed is an arg.'nil

By now, I'm a dead duck.
If I follow the logic, startup-setup-paths would be better calling 
paths-find-lisp-directory BEFORE messing about with the package 
hierarchy - to get the primary lisp directory in place.
And, again with only a cursory examination, paths-find-version-directory 
should be called with envvar="EMACSLOADPATH"

Of course, it could also be made to work if I could figure how to make 
the windows version follow a windows shortcut in "chase-symlink".

Any solution, sadly, touches on a dumped routine.  My score at building 
my own is running a dead 'nil at this time.

=
And, Adrian, thanks greatly for the patience.  I know the answers are 
all here inside the beast -- just waiting to be found by the persistent 
and patient.


-- 
David A. Cobb, Software Engineer, Public Access Advocate
"By God's Grace, I am a Christian man; by my actions a great sinner." -- The Way of a Pilgrim: R.French, Tr.
Life is too short to tolerate crappy software!