Re: master 8fd97b1 1/2: Fix warning generated by indian.el + quail.el

Re: master 8fd97b1 1/2: Fix warning generated by indian.el + quail.el

[Stefan Monnier]

> +  ;; This is a funcall to avoid `quail-update-leim-list-file'
> +  ;; determining that this is a quail definition (it searches for
> +  ;; "(quail-define-package").
> +  (funcall #'quail-define-package pkgname lang title nil docstring

Would the patch below fix this problem more directly?


        Stefan


diff --git a/lisp/international/quail.el b/lisp/international/quail.el
index 5cfa79203d..576189a7de 100644
--- a/lisp/international/quail.el
+++ b/lisp/international/quail.el
@@ -3081,7 +3081,7 @@ quail-update-leim-list-file
 	    (insert-file-contents (car pkg-list))
 	    (goto-char (point-min))
 	    ;; Don't get fooled by commented-out code.
-	    (while (re-search-forward "^[ \t]*(quail-define-package" nil t)
+	    (while (re-search-forward "^[ \t]*(quail-define-package[ \t\n]*\"" nil t)
 	      (goto-char (match-beginning 0))
 	      (condition-case nil
 		  (let ((form (read (current-buffer))))

Re: master 8fd97b1 1/2: Fix warning generated by indian.el + quail.el

[Lars Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> Would the patch below fix this problem more directly?

[...]

> -	    (while (re-search-forward "^[ \t]*(quail-define-package" nil t)
> +	    (while (re-search-forward "^[ \t]*(quail-define-package[ \t\n]*\"" nil t)

I'm not sure?  This all ends up as a (register-input-method ...) call,
which is a function, from which I inferred that literal strings aren't
necessarily required here (i.e., that you can quail-define-package with
a variable name as the first element).  

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: master 8fd97b1 1/2: Fix warning generated by indian.el + quail.el

[Stefan Monnier]

>> Would the patch below fix this problem more directly?
>
> [...]
>
>> -	    (while (re-search-forward "^[ \t]*(quail-define-package" nil t)
>> +	    (while (re-search-forward "^[ \t]*(quail-define-package[ \t\n]*\"" nil t)
>
> I'm not sure?  This all ends up as a (register-input-method ...) call,
> which is a function, from which I inferred that literal strings aren't
> necessarily required here (i.e., that you can quail-define-package with
> a variable name as the first element).  

But IIRC those that we want to extract are specifically those that
are already "fully computed".  In theory we could miss some things like

    (let ((name "foo"))
      (quail-define-package name ...))

but that seems very unlikely.


        Stefan

Re: master 8fd97b1 1/2: Fix warning generated by indian.el + quail.el

[Lars Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> But IIRC those that we want to extract are specifically those that
> are already "fully computed".  In theory we could miss some things like
>
>     (let ((name "foo"))
>       (quail-define-package name ...))
>
> but that seems very unlikely.

It seems kinda fragile either way...  but I have no opinion here,
really.  If you think fixing this via the regexp in quail.el instead is
better as a long-term solution, I have no objections to that.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no