Achim Gratz <strom...@nexgo.de> writes:
> Nicolas Goaziou writes:
>> declarations (instead of requires) are here to avoid circular
>> dependencies. Why do you want to revert that in the first place?
>
> ...because so many of them were missing that it seemed easier to simply
> pull in the definitions by requirement.
But it isn't easier in the end, as you can see.
By the way, I have a (ugly and mostly working) function to add those
declarations.
#+begin_src emacs-lisp
(defun ngz-declare-function (package)
"Insert necessary declare-function calls at point.
PACKAGE is the package from which functions should be declared."
(interactive "sPackage name: ")
(dolist (obj (save-excursion
(goto-char (point-min))
(let (acc (re (format "\\(^\\|[^`]\\)\\(%s-[-a-z]+\\)"
package)))
(while (re-search-forward re nil t)
(let ((func (org-match-string-no-properties 2)))
(unless (member func acc) (push func acc))))
(sort acc 'string<))))
(cond
((fboundp (intern obj))
(let ((args
(let* ((desc (save-window-excursion
(describe-function (intern obj))))
(args
(and (string-match (format "(%s \\([^)]+\\)?)" obj) desc)
(downcase
(replace-regexp-in-string
"[ \t]*\n[ \t]*" " "
(org-match-string-no-properties 1 desc))))))
(insert
(format "(declare-function %s \"%s\" (%s))\n"
obj package (or args ""))))))))
((boundp (intern obj)) (insert (format "(defvar %s)\n" obj))))))
#+end_src
> Yes, that was what I was thinking as well. But that is not actually the
> problem with the compilation, but rather your use of cl macros in some
> places. Somehow this messes up the compilation in a non-obvious way,
> see my other post. If you want to try it yourself, delete one of the
> org-e-*.elc files after the first compilation and try to compile that
> again and watch it explode...
I do not notice anything like this. There are many compilation errors
on some files, but they are the same before and after removing an
org-e-*.el file.
Regards,
