* bug#67523: check-declare doesn't account for shorthands
@ 2023-11-29 8:03 Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors
2023-11-29 9:56 ` João Távora
0 siblings, 1 reply; 5+ messages in thread
From: Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors @ 2023-11-29 8:03 UTC (permalink / raw)
To: 67523; +Cc: Adam Porter, Jonas Bernoulli, João Távora
On Emacs 29.1, when running `check-declare-file' on a file with
`declare-function' forms, I get
Warning (check-declare): said ‘some-nice-string-utils-foobar’ was defined in
some-nice-string-utils.el: function not found
The problem is that `check-declare-verify' attempts to search for the
full symbol name using this regular expression:
(setq re (format (if cflag
"^[ \t]*\\(DEFUN\\)[ \t]*([ \t]*\"%s\""
"^[ \t]*(\\(fset[ \t]+'\\|\
cl-def\\(?:generic\\|method\\|un\\)\\|\
def\\(?:un\\|subst\\|foo\\|method\\|class\\|\
ine-\\(?:derived\\|generic\\|\\(?:global\\(?:ized\\)?-\\)?minor\\)-mode\\|\
\\(?:ine-obsolete-function-\\)?alias[ \t]+'\\|\
ine-overloadable-function\\)\\)\
[ \t]*%s\\([ \t;]+\\|$\\)")
(regexp-opt (mapcar 'cadr fnlist) t)))
(while (re-search-forward re nil t)
...
)
where (mapcar 'cadr fnlist) is the full symbol name.
Since the full symbol name never appears in the file in which it was
defined, re-search-forward never finds it, and so "function not found".
A potential solution could be to convert the longhand symbol into its
shorthand form and pass that into re-search-forward. This is tricky
since there may be multiple different shorthands which could yield the
same longhand form. It might be more feasible to run re-search-forward
on a known common suffix portion of the symbol name, then with point on
the suspected definition, run `intern-soft' to get the full symbol name.
A workaround is to not use shorthands in function definitions.
Thoughts?
Joseph
^ permalink raw reply [flat|nested] 5+ messages in thread
* bug#67523: check-declare doesn't account for shorthands
2023-11-29 8:03 bug#67523: check-declare doesn't account for shorthands Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors
@ 2023-11-29 9:56 ` João Távora
2023-11-29 10:35 ` João Távora
0 siblings, 1 reply; 5+ messages in thread
From: João Távora @ 2023-11-29 9:56 UTC (permalink / raw)
To: Joseph Turner; +Cc: 67523, Adam Porter, Jonas Bernoulli
On Wed, Nov 29, 2023 at 9:12 AM Joseph Turner <joseph@ushin.org> wrote:
> A potential solution could be to convert the longhand symbol into its
> shorthand form and pass that into re-search-forward. This is tricky
> since there may be multiple different shorthands which could yield the
> same longhand form. It might be more feasible to run re-search-forward
> on a known common suffix portion of the symbol name, then with point on
> the suspected definition, run `intern-soft' to get the full symbol name.
No, this is brittle. Check-declare, if it's to be useful (is it?)
is probably meant to be as precise as possible.
> A workaround is to not use shorthands in function definitions.
That's letting the bad guys win :-)
> Thoughts?
As usual, my thoughts are that tools that read Lisp code
should use the Lisp reader, not regular expressions.
Here, check-declare should just walk the whole file.
When it finds a a symbol atom that matches a definition
form, look in the next atom and check if it matches the probe.
If you don't want to intern all the symbols in the you can read
to a separate obarray I think.
João
^ permalink raw reply [flat|nested] 5+ messages in thread
* bug#67523: check-declare doesn't account for shorthands
2023-11-29 9:56 ` João Távora
@ 2023-11-29 10:35 ` João Távora
2023-11-29 11:12 ` João Távora
0 siblings, 1 reply; 5+ messages in thread
From: João Távora @ 2023-11-29 10:35 UTC (permalink / raw)
To: Joseph Turner; +Cc: 67523, Adam Porter, Jonas Bernoulli
On Wed, Nov 29, 2023 at 9:56 AM João Távora <joaotavora@gmail.com> wrote:
>
> On Wed, Nov 29, 2023 at 9:12 AM Joseph Turner <joseph@ushin.org> wrote:
>
> > A potential solution could be to convert the longhand symbol into its
> > shorthand form and pass that into re-search-forward. This is tricky
> > since there may be multiple different shorthands which could yield the
> > same longhand form. It might be more feasible to run re-search-forward
> > on a known common suffix portion of the symbol name, then with point on
> > the suspected definition, run `intern-soft' to get the full symbol name.
>
> No, this is brittle. Check-declare, if it's to be useful (is it?)
> is probably meant to be as precise as possible.
>
> > A workaround is to not use shorthands in function definitions.
>
> That's letting the bad guys win :-)
>
> > Thoughts?
>
> As usual, my thoughts are that tools that read Lisp code
> should use the Lisp reader, not regular expressions.
>
> Here, check-declare should just walk the whole file.
Or maybe just this 100% guaranteed untested patch would work:
diff --git a/lisp/emacs-lisp/check-declare.el b/lisp/emacs-lisp/check-declare.el
index c887d95210c..00440276643 100644
--- a/lisp/emacs-lisp/check-declare.el
+++ b/lisp/emacs-lisp/check-declare.el
@@ -145,6 +145,7 @@ check-declare-verify
(if (file-regular-p fnfile)
(with-temp-buffer
(insert-file-contents fnfile)
+ (hack-local-variables) ;; for shorthands
;; defsubst's don't _have_ to be known at compile time.
(setq re (format (if cflag
"^[ \t]*\\(DEFUN\\)[ \t]*([ \t]*\"%s\""
@@ -158,7 +159,7 @@ check-declare-verify
(regexp-opt (mapcar 'cadr fnlist) t)))
(while (re-search-forward re nil t)
(skip-chars-forward " \t\n")
- (setq fn (match-string 2)
+ (setq fn (symbol-name (car (read-from-string (match-string 2))))
type (match-string 1)
;; (min . max) for a fixed number of arguments, or
;; arglists with optional elements.
^ permalink raw reply related [flat|nested] 5+ messages in thread
* bug#67523: check-declare doesn't account for shorthands
2023-11-29 10:35 ` João Távora
@ 2023-11-29 11:12 ` João Távora
2023-12-10 10:57 ` Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors
0 siblings, 1 reply; 5+ messages in thread
From: João Távora @ 2023-11-29 11:12 UTC (permalink / raw)
To: Joseph Turner; +Cc: 67523, Adam Porter, Jonas Bernoulli
On Wed, Nov 29, 2023 at 10:35 AM João Távora <joaotavora@gmail.com> wrote:
>
> On Wed, Nov 29, 2023 at 9:56 AM João Távora <joaotavora@gmail.com> wrote:
> >
> > On Wed, Nov 29, 2023 at 9:12 AM Joseph Turner <joseph@ushin.org> wrote:
> >
> > > A potential solution could be to convert the longhand symbol into its
> > > shorthand form and pass that into re-search-forward. This is tricky
> > > since there may be multiple different shorthands which could yield the
> > > same longhand form. It might be more feasible to run re-search-forward
> > > on a known common suffix portion of the symbol name, then with point on
> > > the suspected definition, run `intern-soft' to get the full symbol name.
> >
> > No, this is brittle. Check-declare, if it's to be useful (is it?)
> > is probably meant to be as precise as possible.
> >
> > > A workaround is to not use shorthands in function definitions.
> >
> > That's letting the bad guys win :-)
> >
> > > Thoughts?
> >
> > As usual, my thoughts are that tools that read Lisp code
> > should use the Lisp reader, not regular expressions.
> >
> > Here, check-declare should just walk the whole file.
>
> Or maybe just this 100% guaranteed untested patch would work:
Sorry, that was 100% untested indeed. This patch seems to work:
diff --git a/lisp/emacs-lisp/check-declare.el b/lisp/emacs-lisp/check-declare.el
index c887d95210c..bc3844ca9be 100644
--- a/lisp/emacs-lisp/check-declare.el
+++ b/lisp/emacs-lisp/check-declare.el
@@ -145,21 +145,26 @@ check-declare-verify
(if (file-regular-p fnfile)
(with-temp-buffer
(insert-file-contents fnfile)
+ (unless cflag
+ ;; for syntax and shorthands
+ (lisp-data-mode)
+ (hack-local-variables))
;; defsubst's don't _have_ to be known at compile time.
- (setq re (format (if cflag
- "^[ \t]*\\(DEFUN\\)[ \t]*([ \t]*\"%s\""
- "^[ \t]*(\\(fset[ \t]+'\\|\
+ (setq re (if cflag
+ (format "^[ \t]*\\(DEFUN\\)[ \t]*([ \t]*\"%s\""
+ (regexp-opt (mapcar 'cadr fnlist) t))
+ "^[ \t]*(\\(fset[ \t]+'\\|\
cl-def\\(?:generic\\|method\\|un\\)\\|\
def\\(?:un\\|subst\\|foo\\|method\\|class\\|\
ine-\\(?:derived\\|generic\\|\\(?:global\\(?:ized\\)?-\\)?minor\\)-mode\\|\
\\(?:ine-obsolete-function-\\)?alias[ \t]+'\\|\
ine-overloadable-function\\)\\)\
-[ \t]*%s\\([ \t;]+\\|$\\)")
- (regexp-opt (mapcar 'cadr fnlist) t)))
+[ \t]*\\(\\(?:\\sw\\|\\s_\\)+\\)\\([ \t;]+\\|$\\)"))
(while (re-search-forward re nil t)
(skip-chars-forward " \t\n")
- (setq fn (match-string 2)
- type (match-string 1)
+ (setq fn (symbol-name (car (read-from-string (match-string 2)))))
+ (when (member fn (mapcar 'cadr fnlist))
+ (setq type (match-string 1)
;; (min . max) for a fixed number of arguments, or
;; arglists with optional elements.
;; (min) for arglists with &rest.
@@ -202,7 +207,7 @@ check-declare-verify
(t
'err))
;; alist of functions and arglist signatures.
- siglist (cons (cons fn sig) siglist)))))
+ siglist (cons (cons fn sig) siglist))))))
(dolist (e fnlist)
(setq arglist (nth 2 e)
type
^ permalink raw reply related [flat|nested] 5+ messages in thread
* bug#67523: check-declare doesn't account for shorthands
2023-11-29 11:12 ` João Távora
@ 2023-12-10 10:57 ` Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors
0 siblings, 0 replies; 5+ messages in thread
From: Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors @ 2023-12-10 10:57 UTC (permalink / raw)
To: João Távora; +Cc: 67523, Adam Porter, Jonas Bernoulli
João Távora <joaotavora@gmail.com> writes:
> On Wed, Nov 29, 2023 at 10:35 AM João Távora <joaotavora@gmail.com> wrote:
>>
>> On Wed, Nov 29, 2023 at 9:56 AM João Távora <joaotavora@gmail.com> wrote:
>> >
>> > On Wed, Nov 29, 2023 at 9:12 AM Joseph Turner <joseph@ushin.org> wrote:
>> >
>> > > A potential solution could be to convert the longhand symbol into its
>> > > shorthand form and pass that into re-search-forward. This is tricky
>> > > since there may be multiple different shorthands which could yield the
>> > > same longhand form. It might be more feasible to run re-search-forward
>> > > on a known common suffix portion of the symbol name, then with point on
>> > > the suspected definition, run `intern-soft' to get the full symbol name.
>> >
>> > No, this is brittle. Check-declare, if it's to be useful (is it?)
>> > is probably meant to be as precise as possible.
>> >
>> > > A workaround is to not use shorthands in function definitions.
>> >
>> > That's letting the bad guys win :-)
>> >
>> > > Thoughts?
>> >
>> > As usual, my thoughts are that tools that read Lisp code
>> > should use the Lisp reader, not regular expressions.
>> >
>> > Here, check-declare should just walk the whole file.
>>
>> Or maybe just this 100% guaranteed untested patch would work:
>
> Sorry, that was 100% untested indeed. This patch seems to work:
>
> diff --git a/lisp/emacs-lisp/check-declare.el b/lisp/emacs-lisp/check-declare.el
> index c887d95210c..bc3844ca9be 100644
> --- a/lisp/emacs-lisp/check-declare.el
> +++ b/lisp/emacs-lisp/check-declare.el
> @@ -145,21 +145,26 @@ check-declare-verify
> (if (file-regular-p fnfile)
> (with-temp-buffer
> (insert-file-contents fnfile)
> + (unless cflag
> + ;; for syntax and shorthands
> + (lisp-data-mode)
> + (hack-local-variables))
> ;; defsubst's don't _have_ to be known at compile time.
> - (setq re (format (if cflag
> - "^[ \t]*\\(DEFUN\\)[ \t]*([ \t]*\"%s\""
> - "^[ \t]*(\\(fset[ \t]+'\\|\
> + (setq re (if cflag
> + (format "^[ \t]*\\(DEFUN\\)[ \t]*([ \t]*\"%s\""
> + (regexp-opt (mapcar 'cadr fnlist) t))
> + "^[ \t]*(\\(fset[ \t]+'\\|\
> cl-def\\(?:generic\\|method\\|un\\)\\|\
> def\\(?:un\\|subst\\|foo\\|method\\|class\\|\
> ine-\\(?:derived\\|generic\\|\\(?:global\\(?:ized\\)?-\\)?minor\\)-mode\\|\
> \\(?:ine-obsolete-function-\\)?alias[ \t]+'\\|\
> ine-overloadable-function\\)\\)\
> -[ \t]*%s\\([ \t;]+\\|$\\)")
> - (regexp-opt (mapcar 'cadr fnlist) t)))
> +[ \t]*\\(\\(?:\\sw\\|\\s_\\)+\\)\\([ \t;]+\\|$\\)"))
> (while (re-search-forward re nil t)
> (skip-chars-forward " \t\n")
> - (setq fn (match-string 2)
> - type (match-string 1)
> + (setq fn (symbol-name (car (read-from-string (match-string 2)))))
> + (when (member fn (mapcar 'cadr fnlist))
> + (setq type (match-string 1)
> ;; (min . max) for a fixed number of arguments, or
> ;; arglists with optional elements.
> ;; (min) for arglists with &rest.
> @@ -202,7 +207,7 @@ check-declare-verify
> (t
> 'err))
> ;; alist of functions and arglist signatures.
> - siglist (cons (cons fn sig) siglist)))))
> + siglist (cons (cons fn sig) siglist))))))
> (dolist (e fnlist)
> (setq arglist (nth 2 e)
> type
IIUC, this patch is being merged in response to bug#67390.
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2023-12-10 10:57 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2023-11-29 8:03 bug#67523: check-declare doesn't account for shorthands Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors
2023-11-29 9:56 ` João Távora
2023-11-29 10:35 ` João Távora
2023-11-29 11:12 ` João Távora
2023-12-10 10:57 ` Joseph Turner via Bug reports for GNU Emacs, the Swiss army knife of text editors
Code repositories for project(s) associated with this external index
https://git.savannah.gnu.org/cgit/emacs.git
https://git.savannah.gnu.org/cgit/emacs/org-mode.git
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.