GNU bug report logs - #39823
26.3; update-directory-autoloads regression from Emacs 26 to Emacs 27

Previous Next

Package: emacs;

Reported by: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>

Date: Fri, 28 Feb 2020 15:13:02 UTC

Severity: normal

Tags: fixed

Found in version 26.3

Fixed in version 28.1

Done: Lars Ingebrigtsen <larsi <at> gnus.org>

Bug is archived. No further changes may be made.

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 39823 in the body.
You can then email your comments to 39823 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

View this report as an mbox folder, status mbox, maintainer mbox


Report forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Fri, 28 Feb 2020 15:13:02 GMT) Full text and rfc822 format available.

Acknowledgement sent to Maxim Cournoyer <maxim.cournoyer <at> gmail.com>:
New bug report received and forwarded. Copy sent to bug-gnu-emacs <at> gnu.org. (Fri, 28 Feb 2020 15:13:02 GMT) Full text and rfc822 format available.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: bug-gnu-emacs <at> gnu.org
Subject: 26.3; update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Fri, 28 Feb 2020 10:11:54 -0500
Hello!

After passing from Emacs 26 to Emacs 27, the following snippet of code
doesn't behave the same:

--8<---------------cut here---------------start------------->8---
emacs --quick --batch --eval "(progn
 ;(require 'autoload)
 (let ((backup-inhibited t)
       (generated-autoload-file \"/tmp/toto\"))
   (update-directory-autoloads \"/tmp\")))"
--8<---------------cut here---------------end--------------->8---

Works on Emacs 26.3, but fails on Emacs 27.0.50, with the error message:

Wrong type argument: stringp, nil

If uncommenting the require line, like:

--8<---------------cut here---------------start------------->8---
emacs --quick --batch --eval "(progn
 (require 'autoload)
 (let ((backup-inhibited t)
       (generated-autoload-file \"/tmp/toto\"))
   (update-directory-autoloads \"/tmp\")))"
--8<---------------cut here---------------end--------------->8---

It now proceeds with the following output:

  INFO     Scraping files for toto... 
  INFO     Scraping files for toto...done


I've isolated the autoload file containing the
update-directory-autoloads definition to be
/gnu/store/zpmsyn471y4hpgsbz652h4szyskzc2bm-profile/share/emacs/27.0.50/lisp/emacs-lisp/autoload.elc,
using:

--8<---------------cut here---------------start------------->8---
strace emacs --quick --batch --eval "(progn
 ;(require 'autoload)
 (let ((backup-inhibited t)
       (generated-autoload-file \"/tmp/toto\"))
   (update-directory-autoloads \"/tmp\")))" |& grep -E 'autoload.elc.* = 0'
stat("/gnu/store/zpmsyn471y4hpgsbz652h4szyskzc2bm-profile/share/emacs/27.0.50/lisp/emacs-lisp/autoload.elc", {st_mode=S_IFREG|0444, st_size=29029, ...}) = 0--8<---------------cut here---------------end--------------->8---


Its content is copied below:
--8<---------------cut here---------------start------------->8---
cat /gnu/store/zpmsyn471y4hpgsbz652h4szyskzc2bm-profile/share/emacs/27.0.50/lisp/emacs-lisp/autoload.elc
;ELC
;;; Compiled
;;; in Emacs version 27.0.50
;;; with all optimizations.

;;; This file uses dynamic docstrings, first added in Emacs 19.29.

;;; This file does not contain utf-8 non-ASCII characters,
;;; and so can be loaded in Emacs versions earlier than 23.

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;


(byte-code "\300\301!\210\300\302!\207" [require lisp-mode lisp-mnt] 2)
#@481 File into which to write autoload definitions.
A Lisp file can set this in its local variables section to make
its autoloads go somewhere else.

If this is a relative file name, the directory is determined as
follows:
 - If a Lisp file defined `generated-autoload-file' as a
   file-local variable, use its containing directory.
 - Otherwise use the "lisp" subdirectory of `source-directory'.

The autoload file is assumed to contain a trailer starting with a
FormFeed character.
(defvar generated-autoload-file nil (#$ . 484))
(put 'generated-autoload-file 'safe-local-variable 'stringp)
#@270 Load name for `autoload' statements generated from autoload cookies.
If nil, this defaults to the file name, sans extension.
Typically, you need to set this when the directory containing the file
is not in `load-path'.
This also affects the generated cus-load.el file.
(defvar generated-autoload-load-name nil (#$ . 1080))
(put 'generated-autoload-load-name 'safe-local-variable 'stringp)
#@447 Magic comment indicating the following form should be autoloaded.
Used by \[update-file-autoloads].  This string should be
meaningless to Lisp (e.g., a comment).

This string is used:

;;;###autoload
(defun function-to-be-autoloaded () ...)

If this string appears alone on a line, the following form will be
read and an autoload made for it.  If there is further text on the line,
that text will be copied verbatim to `generated-autoload-file'.
(defvar generate-autoload-cookie ";;;###autoload" (#$ . 1476))
#@68 If non-nil, list of absolute file names not to scan for autoloads.
(defvar autoload-excludes nil (#$ . 1991))
#@75 String that marks the form at the start of a new file's autoload section.
(defconst generate-autoload-section-header "\f\n;;;### " (#$ . 2107))
#@72 String which indicates the end of the section of autoloads for a file.
(defconst generate-autoload-section-trailer "\n;;;***\n" (#$ . 2257))
#@64 String to add on each continuation of the section header form.
(defconst generate-autoload-section-continuation ";;;;;; " (#$ . 2404))
#@52 Value to insert when `autoload-timestamps' is nil.
(defconst autoload--non-timestamp '(0 0 0 0) (#$ . 2545))
#@892 Non-nil means insert a timestamp for each input file into the output.
We use these in incremental updates of the output file to decide
if we need to rescan an input file.  If you set this to nil,
then we use the timestamp of the output file instead.  As a result:
 - for fixed inputs, the output will be the same every time
 - incremental updates of the output file might not be correct if:
   i) the timestamp of the output file cannot be trusted (at least
     relative to that of the input files)
   ii) any of the input files can be modified during the time it takes
      to create the output
   iii) only a subset of the input files are scanned
   These issues are unlikely to happen in practice, and would arguably
   represent bugs in the build system.  Item iii) will happen if you
   use a command like `update-file-autoloads', though, since it only
   checks a single input file.
(defvar autoload-timestamps nil (#$ . 2661))
#@343 Turn FORM into an autoload or defvar for source file FILE.
Returns nil if FORM is not a special autoload form (i.e. a function definition
or macro definition or a defcustom).
If EXPANSION is non-nil, we're processing the macro expansion of an
expression, in which case we want to handle forms differently.

(fn FORM FILE &optional EXPANSION)
(defalias 'make-autoload #[770 "C\211\242\242\301\203\207\302=\203\207\242A\211A\211 <at> A\211\303:\203t@\211\304\267\202dA\211:\203\346\211@\211:\203\%\262\202\261\240\f\30106A\211:\203\266\211@\211\306=\203\245A\211\204\224A\211:\203\203\211 <at> A\211\204r\n\307%\202~\n\301%\266\202\202\217
             \301\f%\262\202\301\n\301
\301    %\262\202\361\f\30%\262\202oA\211:\203T\211@\211:\203D\211@\211\306=\2034AA\211\204#
                                                                                            \307%\202/
\301    %\262\202_\f\30%\262\202o                                                                     \301\f%\266\202\202?      \301\n%\262\202O
\"\322\"\323\324%\242A\"\"\211\205\306\325B\262\262\202\372\326>\203\366\327!\203\366\330\242!)\211\262\203\366\211@\331>\203\366\332\333#\202\372\334\267\202\352\335>\242A@\336\267\202\337\2428\202\340\202\301\202\307\341\342\"\206!\343\242\233\211@;\2051\211A\262\242<\203<\344\"\262\345<\203F\202I\305D
           \n\346>\203W\307\202\\@\242\347=     \205b\350\257\266\205\202\372\242A@\337\2428\351\2428\352\305D\305D\n\257\266\203\202\372\242\243\242\242\243\243\242\242\243\243\243\242\325\353F\354\305D
                                                        \3551\262\356
                                                                     \242>A@?0\202\264\210\301FE\266\203\202\372\242A@\357\360\361\305D\362BBDC\363\364\305\nD\365BB\301\366\305D\367\370\305D\371BBFFE\262\202\372\203\371\242:\203\371\242\202\372\301\266\202\207" [load-file-name nil defalias #[1285 "\211\242\300>\205\n\211A@\301:\203Z@\211\302=\203HA\211:\2036\211 <at> A



                                                                        &\266\202\202C                          \303\211&\262\\303\211&\262\202g\211\303\211&\207" ['function #[1542 "\242;\203
                                            @\202\242;\205@:\205l@\211:\203M\211@\211\300=\203+\301\202HA\211:\205F\211@\211:\205D\211@\211\300=\205B\301\262\262\262\262\202jA\211:\205h\211@\211:\205f\211@\211\300=\205d\301\262\262\262\262:\203w\302\"\262\30\242A@   \257\207" [interactive t help-add-fundoc-usage autoload] 14 "\n\n(fn FORM FILE REST TYPE BODY ARGS)"] lambda t] 18 "\n\n(fn FORM FILE REST TYPE FUN)"] #s(hash-table size 2 test eq rehash-size 1.5 rehash-threshold 0.8125 purecopy t data (cons 41 quote 246)) quote macro t (progn prog1) :autoload-end copy-sequence delq mapcar make-byte-code 257 "\302\300\301#\207" vconcat vector [make-autoload] 5 "\n\n(fn FORM)" progn (easy-mmode-define-global-mode define-global-minor-mode define-globalized-minor-mode defun defmacro easy-mmode-define-minor-mode define-minor-mode define-inline cl-defun cl-defmacro cl-defgeneric cl-defstruct pcase-defmacro) macrop macroexpand (progn prog1 defalias) make-autoload expansion #s(hash-table size 17 test eq rehash-size 1.5 rehash-threshold 0.8125 purecopy t data (define-skeleton 508 define-derived-mode 508 define-compilation-mode 508 define-generic-mode 508 easy-mmode-define-global-mode 508 define-global-minor-mode 508 define-globalized-minor-mode 508 easy-mmode-define-minor-mode 508 define-minor-mode 508 cl-defun 508 defun* 508 cl-defmacro 508 defmacro* 508 define-overloadable-function 508 defclass 617 defcustom 646 defgroup 699)) (defmacro cl-defmacro defmacro*) #s(hash-table size 11 test eq rehash-size 1.5 rehash-threshold 0.8125 purecopy t data (define-overloadable-function 521 cl-defmacro 521 cl-defun 521 defmacro* 521 defun* 521 defmacro 521 defun 521 define-skeleton 528 define-compilation-mode 532 define-derived-mode 532 define-generic-mode 532)) 2 (&optional str arg) function-get doc-string-elt 3 help-add-fundoc-usage autoload (define-skeleton define-derived-mode define-generic-mode easy-mmode-define-global-mode define-global-minor-mode define-globalized-minor-mode easy-mmode-define-minor-mode define-minor-mode) interactive 'macro 4 eieio-defclass-autoload defvar custom-autoload (error) :set let loads get ('custom-loads) if member (loads) put 'custom-loads cons (loads)] 29 (#$ . 3604)])
#@72 Visit the autoload file for the current buffer, and return its buffer.
(defalias 'autoload-find-generated-file #[0 "\303\304\305\306 r\307\310!!q\210\311 \312U\203!\313\304\305#c\210p,\207" [delay-mode-hooks enable-local-eval enable-local-variables :safe nil t autoload-generated-file find-file-noselect autoload-ensure-file-writeable buffer-size 0 autoload-rubric] 8 (#$ . 8148)])
#@165 Return `generated-autoload-file' as an absolute name.
If local to the current buffer, expand using the default directory;
otherwise, using `source-directory'/lisp.
\302\304s 'autol\"\"\207" [generated-autoload-file source-directory expand-file-name local-variable-p "lisp"] 5 (#$ . 8546)])
#@174 Read a section header form.
Since continuation lines have been marked as comments,
we must copy the text of the form and remove those comment
markers before we call `read'.
(defalias 'autoload-read-section-header #[0 "\301 \302\303\304\305\306!\307\"\310$\216`\311\312y\210\31!\203\312y\210\202`{\262r\314\315!q\210\316 \210\211c\210eb\210\31\311\320#\203B\321\322!\210\2023eb\210\323p!)\266\202)\207" [generate-autoload-section-continuation match-data make-byte-code 0 "\301\300\302\"\207" vconcat vector [set-match-data evaporate] 3 nil 1 looking-at get-buffer-create " *autoload*" erase-buffer search-forward t replace-match " " read] 7 (#$ . 8897)])
#@56 Buffer which gets the output of `autoload-print-form'.
(defvar autoload-print-form-outbuf nil (#$ . 9570))
#@159 Print FORM such that `make-docfile' will find the docstrings.
The variable `autoload-print-form-outbuf' specifies the buffer to
put the output in.

(fn FORM)
\306\307A\"\207\2119\203\310\207\311\242\312\\203\2548;\203\254S\233\211A\310\241\210\313\314\"\210\315\211\211\211211\203V\211@\316\"\210\313\317\"\210A\266\202\202=\210,\313\320\"\210rq\210`)\313\321\322@!\323\"\"\210rq\210\212\324\325\315#\203\204\323u\210\326c\210\202s*A\204\222\313\327\"\210\202\242\313\317\"\210\313\321\322A!\323\"\"\210\330!\262\266\202\202\271\315\211\211\211331\",\207" [autoload-print-form-outbuf print-escape-nonascii print-quoted print-escape-control-characters print-escape-newlines progn mapcar autoload-print-form nil function-get doc-string-elt princ "\n(" t prin1 " " "\"\\\n" substring prin1-to-string 1 re-search-backward "\n[[(]" "\\" ")" terpri print] 10 (#$ . 9684)])
#@430 Return a string giving the appropriate autoload rubric for FILE.
TYPE (default "autoloads") is a string stating the type of
information contained in FILE.  TYPE "package" acts like the default,
but adds an extra line to the output to modify `load-path'.

If FEATURE is non-nil, FILE will provide a feature.  FEATURE may
be a string naming the feature, otherwise it will be based on
FILE's name.

(fn FILE &optional TYPE FEATURE)
\302\211\262\303\304\206\302\305\205\306\307232\\2054\310\311
                                                             ;\203/
                                                                   \2023\312
                                                                            !\"\313\n\314\260
                                                                                             \207" [file-name-nondirectory "package" "autoloads" ";;; " " --- automatically extracted " "\n;;\n;;; Code:\n\n" "(add-to-list 'load-path (directory-file-name\n                         (or (file-name-directory #$) (car load-path))))\n\n" "\f\n" format "(provide '%s)\n" file-name-sans-extension ";; Local Variables:\n;; version-control: never\n;; no-byte-compile: t\n;; no-update-autoloads: t\n;; coding: utf-8\n;; End:\n;;; " " ends here\n"] 16 (#$ . 10666)])
#@76 Non-nil means `autoload-find-generated-file' makes existing file writable.
(defvar autoload-ensure-writable nil (#$ . 11728))
(put 'autoload-ensure-writable 'risky-local-variable t)
#@13 

(fn FILE)
(defalias 'autoload-ensure-file-writeable #[257 \203)\301!\211\203(\302\303\"\304U\203(\3051#\306\307\303\"\"0\202'\210\202(\210\210\207" [autoload-ensure-writable file-modes logand 128 0 (error) set-file-modes logior] 7 (#$ . 11916)])
#@138 Insert the section-header line,
which lists the file name and which functions are in it, etc.

(fn OUTBUF AUTOLOADS LOAD-NAME FILE TIME)
(defalias 'autoload-insert-section-header #[1285 c\210\302\303!\211\203\304\305!!\202\306\307\25\"\266\310!\210rq\210\212\311y\210l?\205H\312\313!\210\314\315w\210l\204/\316      \261\210\202/*\207" [generate-autoload-section-header generate-autoload-section-continuation getenv "SOURCE_DATE_EPOCH" seconds-to-time string-to-number prin1 autoloads terpri -1 move-to-column 64 "^ \n" nil "\n"] 13 (#$ . 12183)])
#@72 Fetch file and put it in a temp buffer.  Return the buffer.

(fn FILE)
(defalias 'autoload-find-file #[257 "\306!\262r\307\310!q\210\311 \210\312 \210\313\314\315\302!\210\313\316 \210)\317!\320\314\"\210\321\314\322 \210*p)\207" [buffer-undo-list buffer-read-only delay-mode-hooks default-directory enable-local-eval enable-local-variables expand-file-name get-buffer-create " *autoload-file*" kill-all-local-variables erase-buffer t nil make-local-variable emacs-lisp-mode file-name-directory insert-file-contents :safe hack-local-variables] 4 (#$ . 12758)])
#@73 File local variable to prevent scanning this file for autoload cookies.
(defvar no-update-autoloads nil (#$ . 13335))
#@61 Compute the name that will be used to load FILE.

(fn FILE)
\204VA\2045@\262\202%\307\310\311\"!\203L\310\211A\262\242\"\262\202%\312\313\314#\262\202%\315\316\"\203f\317\320\211\224#\202g\207" [default-value generated-autoload-file file-relative-name file-name-directory nil directory-file-name file-name-nondirectory file-exists-p expand-file-name "subdirs.el" mapconcat identity "/" string-match "\\.elc?\\(\\.\\|\\'\\)" substring 0] 9 (#$ . 13459)])
#@315 Insert at point a loaddefs autoload section for FILE.
Autoloads are generated for defuns and defmacros in FILE
marked by `generate-autoload-cookie' (which see).
If FILE is being visited in a buffer, the contents of the buffer
are used.
Return non-nil in the case where no autoloads were added at point.

(fn FILE)
(defalias 'generate-file-autoloads #[257 \302p\")\207" [buffer-file-name generated-autoload-file autoload-generate-file-autoloads] 4 (#$ . 14064) "fGenerate autoloads for file: "])
#@276 If non-nil, autoload will add code to register the prefixes used in a file.
Standard prefixes won't be registered anyway.  I.e. if a file "foo.el" defines
variables or functions that use "foo-" as prefix, that will not be registered.
But all other prefixes will be included.
(defvar autoload-compute-prefixes t (#$ . 14570))
(put 'autoload-compute-prefixes 'safe 'booleanp)
#@305 Target length of the list of definition prefixes per file.
If set too small, the prefixes will be too generic (i.e. they'll use little
memory, we'll end up looking in too many files when we need a particular
prefix), and if set too large, they will be too specific (i.e. they will
cost more memory use).
(defconst autoload-def-prefixes-max-entries 5 (#$ . 14951))
#@100 Target size of definition prefixes.
Don't try to split prefixes that are already longer than that.
(defconst autoload-def-prefixes-max-length 12 (#$ . 15322))
(require 'radix-tree)
#@18 

(fn DEFS FILE)
(defalias 'autoload--make-defs-autoload #[514 \211\203\211@\301\302#\262A\266\202\202\210\211\262\303C\304\305\306\307\310\311!\312\"\313\314%\"\210\211\242\303\240\210\211\203}\211@\211@\211G\315V\203H\211\316\230\203W\317\320\"\204W\321A\322\"\203a\242B\240\210\202u\304A\305\306\323\310\311\\"\312\"\313\314%\"\210\210A\266\202\2023\210\211\242\205\247\324\305\325\326\310\31!\327\"\330\331%\242\"\332\333\334\335\336\337\30\"\340\"DEE\262\207" [radix-tree-empty radix-tree-insert t nil radix-tree-iter-subtrees make-byte-code 514 "\300B\300\242B\240\207" vconcat vector #1=[] 5 "\n\n(fn PREFIX SUBTREE)" 2 "def" string-match ".[[:punct:]]\\'" radix-tree-lookup "" "\300\301PB\300\242B\240\207" mapcar 257 "\211@\211G\301V\204\211G\301=\203302\303\"\203211\202<\304C\305A\306\307\310\311\3\"\313\"\314\315%\"\210\316\317\300\242$\210\304\262\207" [2 string-match "[[:punct:]]" nil radix-tree-iter-mappings make-byte-code 514 "\301\300P\301\242B\240\207" vconcat vector #1# 5 "\n\n(fn S _)" message "Not registering prefix \"%s\" from %s.  Affects: %S"] 12 "\n\n(fn X)" if (fboundp 'register-definition-prefixes) register-definition-prefixes quote sort delq string<] 16 (#$ . 15509)])
#@42 

(fn OTHERBUF OUTBUF ABSFILE LOAD-NAME)
(defalias 'autoload--setup-output #[1028 "\20\206\300\"\206\301\302\"r\211q\210\303 )\207" [autoload-find-destination throw done point-marker] 7 (#$ . 16789)])
#@36 

(fn OUTPUT-START LOAD-NAME FILE)
(defalias 'autoload--print-cookie-text #[771 "\303!\304 !\210\305\306w\210l\203C\30715\310p!n\204\311y\210\312\"\211\204)\26\313!)\266\2020\202Z\314\315\316e`\"$\262\202Z\317\320\306x\210`Tf\321U\203S\311u\210`\311y\210`{!)\207" [standard-output generate-autoload-cookie autoload-print-form-outbuf marker-buffer search-forward "      " nil (debug error) read 1 make-autoload autoload-print-form message "Autoload cookie error in %s:%s %S" count-lines princ " \f    " 32] 10 (#$ . 17008)])
(defvar autoload-builtin-package-versions nil)
#@164 List of strings naming definitions to ignore for prefixes.
More specifically those definitions will not be considered for the
`register-definition-prefixes' call.
(defvar autoload-ignored-definitions '("define-obsolete-function-alias" "define-obsolete-variable-alias" "define-category" "define-key" "defgroup" "defface" "defadvice" "def-edebug-spec" "define-widget" "define-erc-module" "define-erc-response-handler" "defun-rcirc-command") (#$ . 17600))
#@758 Insert an autoload section for FILE in the appropriate buffer.
Autoloads are generated for defuns and defmacros in FILE
marked by `generate-autoload-cookie' (which see).
If FILE is being visited in a buffer, the contents of the buffer are used.
OUTBUF is the buffer in which the autoload statements should be inserted.
If OUTBUF is nil, it will be determined by `autoload-generated-file'.

If provided, OUTFILE is expected to be the file name of OUTBUF.
If OUTFILE is non-nil and FILE specifies a `generated-autoload-file'
different from OUTFILE, then OUTBUF is ignored.

Return non-nil if and only if FILE adds no autoloads to OUTFILE
(or OUTBUF if OUTFILE is nil).  The actual return value is
FILE's modification time.

(fn FILE &optional OUTBUF OUTFILE)
(defalias 'autoload-generate-file-autoloads #[769 "\3061<\307\211\211\211\310!\307\311  !\307\211\3122,r\206(\313\f!q\210
\202D\316!\262          \203g@\317>\203\\       \227\320 \227\232\202a  \320 \232\204g\321\262\212\214~\210A\203\315\322\323!\307\203\313\3241\203\325!0\202\205\210\307\211\262\203\313\322\326!\206\227\327\330!!\211\262\203\313\331$\211\262\203\313\332!\321BC\333\334\335\336\337!BDD\340BB!\210\347y\210\202\327E\350\232\203*\343\351!\203*\352\347!F\235\204*\353\354!B\262\355\347!\210\347y\210\202\327*G\203\270\203\270\356\n\"\211\203\267\$)\332!H\357!\210r\332!q\210\212\211b\210\362!\363\332!\364I\203\245\365
                                                                        !\3668\262\202\247J%\210\367\341\261\266)Kc\210*\210\210\211\203\205\301pr\\"\210\204$\375p!\210)\211?\206+0\2057\365!\3668\262+\266\2110\207\376\377 <at> A$\207" [float-output-format print-level print-length no-update-autoloads noninteractive generated-autoload-load-name (error) nil get-file-buffer expand-file-name done autoload-find-file message "Generating autoloads for %s..." autoload-file-load-name (ms-dos windows-nt) autoload-generated-file t lm-header "version" (error) version-to-list "package" file-name-sans-extension file-name-nondirectory autoload--setup-output marker-buffer princ push purecopy quote intern (package--builtin-versions) "\n" "      \n\f" looking-at regexp-quote autoload--print-cookie-text 59 1 ";;;###autoload" "(\\(def[^ ]+\\) ['(]*\\([^' ()\"\n]+\\)[\n        ]" match-string match-string-no-properties 2 forward-sexp autoload--make-defs-autoload autoload-print-form default-value generated-autoload-file file-relative-name autoload-insert-section-header "actual autoloads are elsewhere" file-attributes 5 ";;; Generated autoloads from " cl--assertion-failed (> (point) output-start) md5 emacs-mule-unix "Generating autoloads for %s...done" kill-buffer error "%s:0:0: error: %s: %s" system-type autoload-builtin-package-versions print-quoted standard-output autoload-excludes generate-autoload-cookie autoload-ignored-definitions autoload-compute-prefixes autoload-print-form-outbuf autoload-timestamps autoload--non-timestamp generate-autoload-section-trailer] 24 (#$ . 18060)])
#@46 Save current buffer to its file, atomically.
(defalias 'autoload--save-buffer #[0 "\304\305  !\306 \307\310\"\307\311        !\206\312\"\313\314\315\316\31!\320\"\321$\nBU\204/\322\"\210\323ed\324\325%\210\326 \210\327      \330#\210*\266\331\324!\210\332 \210
                                                                    \206S\333\334       \"\207" [version-control buffer-file-name kill-emacs-hook noninteractive never make-temp-file default-file-modes logand 384 file-modes 438 make-byte-code 0 "\3011 \302\300!0\207\210\303\207" vconcat vector [(error) delete-file nil] 2 set-file-modes write-region nil 1 backup-buffer rename-file t set-buffer-modified-p set-visited-file-modtime message "Wrote %s"] 10 (#$ . 21513)])
(defalias 'autoload-save-buffers #[0 \205\211A\242q\210\301 \210)\202\207" [autoload-modified-buffers autoload--save-buffer] 2])
#@491 Update the autoloads for FILE.
If prefix arg SAVE-AFTER is non-nil, save the buffer too.

If FILE binds `generated-autoload-file' as a file-local variable,
autoloads are written into that file.  Otherwise, the autoloads
file is determined by OUTFILE.  If called interactively, prompt
for OUTFILE; if called from Lisp with OUTFILE nil, use the
existing value of `generated-autoload-file'.

Return FILE if there was no autoload cookie in it, else nil.

(fn FILE &optional SAVE-AFTER OUTFILE)
(defalias 'update-file-autoloads #[769 "\211\20\303\304\305!    \203203&\306 \210\202&\307\310!\203&\311\312\"\210\211\205++\207" [generated-autoload-file autoload-modified-buffers autoload-timestamps nil t autoload-generate-file-autoloads autoload-save-buffers called-interactively-p interactive message "Autoload section for %s is up to date."] 7 (#$ . 22321) (byte-code "\301\302\301\303!E\207" [current-prefix-arg read-file-name "Update autoloads for file: " "Write autoload definitions to file: "] 4)])
#@360 Find the destination point of the current buffer's autoloads.
FILE is the file name of the current buffer.
LOAD-NAME is the name as it appears in the output.
Returns a buffer whose point is placed at the requested location.
Returns nil if the file's autoloads are up-to-date, otherwise
removes any prior now out-of-date autoload entries.

(fn FILE LOAD-NAME)
(defalias 'autoload-find-destination #[514 "\3052\360\205\n\211\306 \307!\205\310!\3118\262\312r\313 q\210\314  !\315=\204,\316\317!\210\320 \315V\2048\321\32\"\210\32!\204C\321\32\"\210~\210eb\210\211\204\324\325\n\312\326#\203\324\327 \3308\230\203\300\315\224\3318\310\n!\3118\262\203y\33!\204\264\203\216\326
                  D\235\203\216\333\"?\202\251<\203\233\333\"?\202\251;\203\264\334\n\312\211\335$\232\203\264\336\305\312\"\210\202\273\337!\210\326\262\266\202\320\3308\231\203\320\315\224b\210\326\262\210\202H\211\204\341db\210\340\341\312\326#\210p\f>\204\353p\fBp)\266\2050\207" [buffer-file-name buffer-file-coding-system generate-autoload-section-header autoload--non-timestamp autoload-modified-buffers up-to-date autoload-generated-file file-exists-p file-attributes 5 nil autoload-find-generated-file coding-system-eol-type 0 set-buffer-file-coding-system unix buffer-size error "Autoloads file %s lacks boilerplate" file-writable-p "Autoloads file %s is not writable" search-forward t autoload-read-section-header 2 4 buffer-modified-p time-less-p md5 emacs-mule throw autoload-remove-section search-backward "\f"] 17 (#$ . 23343)])
#@14 

(fn BEGIN)
(defalias 'autoload-remove-section #[257 "\211b\210\30!\210\211`|\207" [generate-autoload-section-trailer search-forward] 3 (#$ . 24914)])
#@757 Update autoload definitions for Lisp files in the directories DIRS.
In an interactive call, you must give one argument, the name of a
single directory.  In a call from Lisp, you can supply multiple
directories as separate arguments, but this usage is discouraged.

The function does NOT recursively descend into subdirectories of the
directory or directories specified.

In an interactive call, prompt for a default output file for the
autoload definitions, and temporarily bind the variable
`generated-autoload-file' to this value.  When called from Lisp,
use the existing value of `generated-autoload-file'.  If any Lisp
file binds `generated-autoload-file' as a file-local variable,
write its autoloads into the specified file instead.

(fn &rest DIRS)
(defalias 'update-directory-autoloads #[128 "\306\307 \211\203\211@\310\311\"\204\211B\262A\266\202\202\210\312\313\314\"\315Q\262\316\317\320\321\322\323\324\32!\326\"\327\330%\"\"\306\211\211\211\211\331\332!\203K\333\334!\202L      \335    !\205[\336      !\3278\262r\337 q\210\212\340\341\n!\320\341       \"\"\262eb\210\342
                                  \306\314#\2034\343 \3448\211:\203\321\211@;\203\321\345\346\224!\210\3478\262\314\fD\235\203\235\262\211\211\203\315\211@\336!\3278\262\211\203\305\35\"\204\305B\262\340\f\"\262
                                                                \210A\266\202\202\236\210\202\"\211;\203\"\335!\203\343\211\235\203\356\314\262\345\346\224!\210\202\"\350\3478\211\314\fD\235\203\376\202\377\211\262\336!\3278\262\"\203\"\314\262\345\346\224!\210\351p\n#\203\"\211B\262\211B\262\340     \"\26\266\202p)\206:\352\353\354\355\341        !P!\346 G\306\356%\346\306      \211\203\227\211 <at> T\211\262\306\247\203f <at> Y\203l\357#\210\266\351\306\n#\211\262\203\215\21B\26\350\"\203\220\262\202\220\314\262\nA\266\202\202M\210\360!\210\203\306\361\362\"\262db\210\363\364\306\314#\210\\203\27\202\300\f%\2109c\210\266\204\323\366\306!\210\202\326\367 \210\370 +\207" [autoload-modified-buffers generated-autoload-file buffer-file-name generate-autoload-section-header autoload--non-timestamp autoload-timestamps nil get-load-suffixes string-match "\\.\\(elc\\|so\\|dll\\)" "^[^=.].*" regexp-opt t "\\'" apply nconc mapcar make-byte-code 257 "\301\302!\303\300#\207" vconcat vector [directory-files expand-file-name t] 5 "\n\n(fn DIR)" called-interactively-p interactive read-file-name "Write autoload definitions to file: " file-exists-p file-attributes autoload-find-generated-file delete file-relative-name search-forward autoload-read-section-header 3 autoload-remove-section 0 4 time-less-p autoload-generate-file-autoloads (0 0 0 0) make-progress-reporter byte-compile-info-string "Scraping files for " 10 progress-reporter-do-update progress-reporter-done sort string< search-backward "\f" autoload-insert-section-header set-buffer-modified-p autoload--save-buffer autoload-save-buffers generate-autoload-section-trailer] 21 (#$ . 25075) "DUpdate autoloads from directory: "])
#@191 Update loaddefs.el autoloads in batch mode.
Calls `update-directory-autoloads' on the command line arguments.
Definitions are written to `generated-autoload-file' (which
should be non-nil).
(defalias 'batch-update-autoloads #[0 \204V\304 !\305\306\307!\203S\310\311!r\211q\210\312\313\314\315\316!\317\"\320$\216\321\307!\210\322\323\305\324#\203Q\325\326!\262\327\330\"\204A\331\332\"\262\327\333\"\204'\334B\202'*\210)\266
                                                                                       \305\335\336\"\207" [autoload-excludes generated-autoload-file default-directory command-line-args-left file-name-directory nil file-readable-p "loadup.el" generate-new-buffer " *temp*" make-byte-code 0 "\301\300!\205      \302\300!\207" vconcat vector [buffer-name kill-buffer] 2 insert-file-contents re-search-forward "^(load \"\\([^\"]+\\)\"" t match-string 1 string-match "\\.el\\'" format "%s.el" "\\`site-" expand-file-name apply update-directory-autoloads] 9 (#$ . 28083)])
(provide 'autoload)
--8<---------------cut here---------------end--------------->8---

Could it be a byte compilation bug?

Thanks for the great editor :-)

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Fri, 13 Mar 2020 00:44:01 GMT) Full text and rfc822 format available.

Message #8 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Noam Postavsky <npostavs <at> gmail.com>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: , 39823 <at> debbugs.gnu.org
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Thu, 12 Mar 2020 20:43:09 -0400
[Message part 1 (text/plain, inline)]
Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:

> After passing from Emacs 26 to Emacs 27, the following snippet of code
> doesn't behave the same:
>
> emacs --quick --batch --eval "(progn
>  ;(require 'autoload)
>  (let ((backup-inhibited t)
>        (generated-autoload-file \"/tmp/toto\"))
>    (update-directory-autoloads \"/tmp\")))"
>
>
> Works on Emacs 26.3, but fails on Emacs 27.0.50, with the error message:
>
> Wrong type argument: stringp, nil

The difference is that --eval now evaluates with lexical-binding
enabled, so the let-binding of generated-autoload-file is lexical unless
autoload.el, with its (defvar generated-autoload-file nil ...), has been
loaded.  Another way to demonstrate this:

    emacs -Q --batch -l bug-39823-autoload-regression.el

fails with the same error, with bug-39823-autoload-regression.el as the
attached file in both Emacs 26 and 27.  Remove the lexical-binding
setting, and it succeeds in both.

[bug-39823-autoload-regression.el (text/plain, attachment)]
[Message part 3 (text/plain, inline)]
See also https://lists.gnu.org/r/emacs-devel/2020-03/msg00173.html for
(much) more discussion on this.




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Fri, 13 Mar 2020 08:31:02 GMT) Full text and rfc822 format available.

Message #11 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Noam Postavsky <npostavs <at> gmail.com>
Cc: , 39823 <at> debbugs.gnu.org, maxim.cournoyer <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Fri, 13 Mar 2020 10:30:46 +0200
> From: Noam Postavsky <npostavs <at> gmail.com>
> Date: Thu, 12 Mar 2020 20:43:09 -0400
> Cc: , 39823 <at> debbugs.gnu.org
> 
> > Works on Emacs 26.3, but fails on Emacs 27.0.50, with the error message:
> >
> > Wrong type argument: stringp, nil
> 
> The difference is that --eval now evaluates with lexical-binding
> enabled, so the let-binding of generated-autoload-file is lexical unless
> autoload.el, with its (defvar generated-autoload-file nil ...), has been
> loaded.
> [...]
> See also https://lists.gnu.org/r/emacs-devel/2020-03/msg00173.html for
> (much) more discussion on this.

The "wrap with eval" trick suggested by Stefan in the latter
discussion does work here:

 emacs --quick --batch --eval "(eval '(progn
   (let ((backup-inhibited t)
         (generated-autoload-file \"/tmp/toto\"))
     (update-directory-autoloads \"/tmp\"))))"

  =|
  INFO     Scraping files for toto...75%
  INFO     Scraping files for toto...done




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Fri, 13 Mar 2020 23:02:01 GMT) Full text and rfc822 format available.

Message #14 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Michael Heerdegen <michael_heerdegen <at> web.de>
To: Noam Postavsky <npostavs <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org, Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Subject: Re: bug#39823: 26.3; update-directory-autoloads regression from
 Emacs 26 to Emacs 27
Date: Sat, 14 Mar 2020 00:01:49 +0100
Noam Postavsky <npostavs <at> gmail.com> writes:

> The difference is that --eval now evaluates with lexical-binding
> enabled, so the let-binding of generated-autoload-file is lexical unless
> autoload.el, with its (defvar generated-autoload-file nil ...), has been
> loaded.

For a general solution, would it make sense to let emacs -Q know in
advance which variables are dynamic, by generating some list when
building?

Michael.




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Sat, 14 Mar 2020 02:04:02 GMT) Full text and rfc822 format available.

Message #17 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Noam Postavsky <npostavs <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Fri, 13 Mar 2020 22:03:36 -0400
Hello!

Noam Postavsky <npostavs <at> gmail.com> writes:

> Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:
>
>> After passing from Emacs 26 to Emacs 27, the following snippet of code
>> doesn't behave the same:
>>
>> emacs --quick --batch --eval "(progn
>>  ;(require 'autoload)
>>  (let ((backup-inhibited t)
>>        (generated-autoload-file \"/tmp/toto\"))
>>    (update-directory-autoloads \"/tmp\")))"
>>
>>
>> Works on Emacs 26.3, but fails on Emacs 27.0.50, with the error message:
>>
>> Wrong type argument: stringp, nil
>
> The difference is that --eval now evaluates with lexical-binding
> enabled, so the let-binding of generated-autoload-file is lexical unless
> autoload.el, with its (defvar generated-autoload-file nil ...), has been
> loaded.  Another way to demonstrate this:
>
>     emacs -Q --batch -l bug-39823-autoload-regression.el
>
> fails with the same error, with bug-39823-autoload-regression.el as the
> attached file in both Emacs 26 and 27.  Remove the lexical-binding
> setting, and it succeeds in both.

Thank you for opening my eyes on this; I was scratching my head a bit.

> See also https://lists.gnu.org/r/emacs-devel/2020-03/msg00173.html for
> (much) more discussion on this.

Eh, indeed, that was a bit much :-).  I have nothing against lexcal
scoping (all to the contrary), and always found it strange that I had to
dynamically bound values variable calling `update-directory-autoloads'.
Perhaps a simple fix would be to allow passing the values it needs as
arguments (while maintaining backward compatibility, i.e., make those
arguments optional?).

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Sat, 14 Mar 2020 02:07:01 GMT) Full text and rfc822 format available.

Message #20 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 39823 <at> debbugs.gnu.org, Noam Postavsky <npostavs <at> gmail.com>
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Fri, 13 Mar 2020 22:06:26 -0400
Hello again,

Eli Zaretskii <eliz <at> gnu.org> writes:

>> From: Noam Postavsky <npostavs <at> gmail.com>
>> Date: Thu, 12 Mar 2020 20:43:09 -0400
>> Cc: , 39823 <at> debbugs.gnu.org
>> 
>> > Works on Emacs 26.3, but fails on Emacs 27.0.50, with the error message:
>> >
>> > Wrong type argument: stringp, nil
>> 
>> The difference is that --eval now evaluates with lexical-binding
>> enabled, so the let-binding of generated-autoload-file is lexical unless
>> autoload.el, with its (defvar generated-autoload-file nil ...), has been
>> loaded.
>> [...]
>> See also https://lists.gnu.org/r/emacs-devel/2020-03/msg00173.html for
>> (much) more discussion on this.
>
> The "wrap with eval" trick suggested by Stefan in the latter
> discussion does work here:
>
>  emacs --quick --batch --eval "(eval '(progn
>    (let ((backup-inhibited t)
>          (generated-autoload-file \"/tmp/toto\"))
>      (update-directory-autoloads \"/tmp\"))))"

Oh, that's an interesting trick.  I had not realized `eval' could be
switched to either lexical or dynamic (the default).

Thanks!

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Sat, 14 Mar 2020 03:35:02 GMT) Full text and rfc822 format available.

Message #23 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Richard Stallman <rms <at> gnu.org>
To: Noam Postavsky <npostavs <at> gmail.com>
Cc: , 39823 <at> debbugs.gnu.org, maxim.cournoyer <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Fri, 13 Mar 2020 23:34:48 -0400
[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

Is there any real use for lexical scoping in --eval?
I suggest reverting the default for --eval to dynamic.
We could introduce --evallex to specify lexical scoping
and ask people to tell us if they find it useful.

-- 
Dr Richard Stallman
Chief GNUisance of the GNU Project (https://gnu.org)
Founder, Free Software Foundation (https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)






Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Tue, 17 Mar 2020 02:17:02 GMT) Full text and rfc822 format available.

Message #26 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Richard Stallman <rms <at> gnu.org>
Cc: 39823 <at> debbugs.gnu.org, Noam Postavsky <npostavs <at> gmail.com>
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Mon, 16 Mar 2020 22:16:04 -0400
Hello,

Richard Stallman <rms <at> gnu.org> writes:

> [[[ To any NSA and FBI agents reading my email: please consider    ]]]
> [[[ whether defending the US Constitution against all enemies,     ]]]
> [[[ foreign or domestic, requires you to follow Snowden's example. ]]]
>
> Is there any real use for lexical scoping in --eval?
> I suggest reverting the default for --eval to dynamic.
> We could introduce --evallex to specify lexical scoping
> and ask people to tell us if they find it useful.

Not a bad suggestion, but I actually embrace the change to lexical mode
everywhere (I find it more forward looking, and like that it is
consistently the default everywhere).  I believe the reason for this bug
report is two-fold:

1) I wasn't aware of the change to --eval (my bad -- I don't follow
Emacs development very closely).
2) The API of some older code (such as the `update-directory-autoloads'
procedure from the autoload library, as I found out) requires the use of
dynamic evaluation for its API to function.

Now that I know about 1), it's rather easy to work around 2).  And I'd
guess that with Elisp finally preferring lexical scope by default, 2)
will tend to disappear over time.

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Tue, 17 Mar 2020 11:49:02 GMT) Full text and rfc822 format available.

Message #29 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Noam Postavsky <npostavs <at> gmail.com>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Tue, 17 Mar 2020 07:48:47 -0400
Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:

> I have nothing against lexcal scoping (all to the contrary), and
> always found it strange that I had to dynamically bound values
> variable calling `update-directory-autoloads'.  Perhaps a simple fix
> would be to allow passing the values it needs as arguments (while
> maintaining backward compatibility, i.e., make those arguments
> optional?).

Hmm, I guess we could add a &key style arg (I don't think plain
&optional would be possible in a backward compatible way because the
current arg list is &rest DIRS; there wouldn't be any way to distinguish
the optional directory argument).




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Thu, 19 Mar 2020 01:47:02 GMT) Full text and rfc822 format available.

Message #32 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Richard Stallman <rms <at> gnu.org>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org, npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Wed, 18 Mar 2020 21:46:17 -0400
[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

I still see no way in which lexical as default makes any sense for --eval.

-- 
Dr Richard Stallman
Chief GNUisance of the GNU Project (https://gnu.org)
Founder, Free Software Foundation (https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)






Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Wed, 08 Apr 2020 03:56:02 GMT) Full text and rfc822 format available.

Message #35 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Richard Stallman <rms <at> gnu.org>
Cc: 39823 <at> debbugs.gnu.org, npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Tue, 07 Apr 2020 23:55:18 -0400
Hello Richard,

Richard Stallman <rms <at> gnu.org> writes:

[...]

> I still see no way in which lexical as default makes any sense for --eval.

Why would it be different than for M-: (eval-expression)? 

It seem to be mostly a matter of consistency (moving to lexical scope by
default, everywhere).

For the record, this change of behavior was mitigated in GNU Guix with
the following change, which was one of the suggestions made here (wrap
in eval):
https://git.savannah.gnu.org/cgit/guix.git/commit/?id=afc6b1c0b635e3268795c0f766be408c5e9858e7.

I think it could be nice to update update-directory-autoloads to work in
a lexical setting before closing this bug, though.

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Wed, 08 Apr 2020 06:32:02 GMT) Full text and rfc822 format available.

Message #38 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Wed, 08 Apr 2020 09:31:32 +0300
> From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
> Date: Tue, 07 Apr 2020 23:55:18 -0400
> Cc: 39823 <at> debbugs.gnu.org, npostavs <at> gmail.com
> 
> I think it could be nice to update update-directory-autoloads to work in
> a lexical setting before closing this bug, though.

Update how?

This discussion mentioned a way to turn off lexical-binding in --eval
command-line arguments (which is also in NEWS), so why do we need to
do anything else specifically for update-directory-autoloads?

IOW, I think this bug can be closed now.




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Wed, 08 Apr 2020 16:42:02 GMT) Full text and rfc822 format available.

Message #41 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Wed, 08 Apr 2020 12:41:26 -0400
Hi Eli,

Eli Zaretskii <eliz <at> gnu.org> writes:

>> From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
>> Date: Tue, 07 Apr 2020 23:55:18 -0400
>> Cc: 39823 <at> debbugs.gnu.org, npostavs <at> gmail.com
>> 
>> I think it could be nice to update update-directory-autoloads to work in
>> a lexical setting before closing this bug, though.
>
> Update how?

Update `update-directory-autoloads' so that it accepts a &key style arg
for the ‘generated-autoload-file’ file name.  Currently it must be
dynamically bound prior to calling `update-directory-autoloads'.  This
was discussed earlier in this thread (see:
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39823#29).

> This discussion mentioned a way to turn off lexical-binding in --eval
> command-line arguments (which is also in NEWS), so why do we need to
> do anything else specifically for update-directory-autoloads?

Because I suspect that 'update-directory-autoloads' is probably used
most of the time with the command line --eval (in Makefiles and build
scripts), and that the wrap in eval workaround, while it works, is
rather obscure.

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Wed, 08 Apr 2020 19:10:02 GMT) Full text and rfc822 format available.

Message #44 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Wed, 08 Apr 2020 22:09:13 +0300
> From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
> Cc: 39823 <at> debbugs.gnu.org,  rms <at> gnu.org,  npostavs <at> gmail.com
> Date: Wed, 08 Apr 2020 12:41:26 -0400
> 
> >> I think it could be nice to update update-directory-autoloads to work in
> >> a lexical setting before closing this bug, though.
> >
> > Update how?
> 
> Update `update-directory-autoloads' so that it accepts a &key style arg
> for the ‘generated-autoload-file’ file name.  Currently it must be
> dynamically bound prior to calling `update-directory-autoloads'.

I don't think this is related to this bug report.  It's a separate
issue: you want to enhance the API of update-directory-autoloads.

> > This discussion mentioned a way to turn off lexical-binding in --eval
> > command-line arguments (which is also in NEWS), so why do we need to
> > do anything else specifically for update-directory-autoloads?
> 
> Because I suspect that 'update-directory-autoloads' is probably used
> most of the time with the command line --eval (in Makefiles and build
> scripts), and that the wrap in eval workaround, while it works, is
> rather obscure.

The Emacs build also uses it, it just uses "-l autoload" before --eval.
It sounds like the right solution to me, something that people should
use if they need to call this function from a Makefile.




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Sun, 19 Apr 2020 03:14:02 GMT) Full text and rfc822 format available.

Message #47 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Sat, 18 Apr 2020 23:13:16 -0400
Eli Zaretskii <eliz <at> gnu.org> writes:

>> From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
>> Cc: 39823 <at> debbugs.gnu.org,  rms <at> gnu.org,  npostavs <at> gmail.com
>> Date: Wed, 08 Apr 2020 12:41:26 -0400
>> 
>> >> I think it could be nice to update update-directory-autoloads to work in
>> >> a lexical setting before closing this bug, though.
>> >
>> > Update how?
>> 
>> Update `update-directory-autoloads' so that it accepts a &key style arg
>> for the ‘generated-autoload-file’ file name.  Currently it must be
>> dynamically bound prior to calling `update-directory-autoloads'.

I investigated that path, but I fail to see how that could be done
safely. &key is a Common Lisp thing that is made available in Emacs Lisp
with `cl-defun'. The original signature of `update-directory-autoloads'
file is:

(defun update-directory-autoloads (&rest dirs)

I don't see how &key can be combined with &rest without potentially
causing problems.

Feel free to close this issue.

Thanks,

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Sun, 19 Apr 2020 12:20:01 GMT) Full text and rfc822 format available.

Message #50 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Noam Postavsky <npostavs <at> gmail.com>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: Eli Zaretskii <eliz <at> gnu.org>, 39823 <at> debbugs.gnu.org, rms <at> gnu.org
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Sun, 19 Apr 2020 08:19:06 -0400
Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:

> (defun update-directory-autoloads (&rest dirs)
>
> I don't see how &key can be combined with &rest without potentially
> causing problems.

You can check if the first arg is a keyword like
:generated-autoload-file (which can't be a directory, since it's not a
string), and if it is, then handle the second arg specially.





Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Mon, 22 Jun 2020 04:11:01 GMT) Full text and rfc822 format available.

Message #53 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Noam Postavsky <npostavs <at> gmail.com>
Cc: Eli Zaretskii <eliz <at> gnu.org>, 39823 <at> debbugs.gnu.org, rms <at> gnu.org
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Mon, 22 Jun 2020 00:10:04 -0400
[Message part 1 (text/plain, inline)]
Hello Noam,

Noam Postavsky <npostavs <at> gmail.com> writes:

> Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:
>
>> (defun update-directory-autoloads (&rest dirs)
>>
>> I don't see how &key can be combined with &rest without potentially
>> causing problems.
>
> You can check if the first arg is a keyword like
> :generated-autoload-file (which can't be a directory, since it's not a
> string), and if it is, then handle the second arg specially.

I've given this idea a try with the attached patch.  For some reason I
can't seem to simply eval the definition to test it (Debugger
entered--Lisp error: (void-function byte-compile-info-string).  I'm
rebuilding Emacs with the change right now.

Consider it untested for now.

[0001-lisp-autoload.el-Add-a-keyword-argument-to-update-di.patch (text/x-patch, attachment)]
[Message part 3 (text/plain, inline)]
Thank you!

Maxim

Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Mon, 22 Jun 2020 15:08:01 GMT) Full text and rfc822 format available.

Message #56 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Noam Postavsky <npostavs <at> gmail.com>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: Eli Zaretskii <eliz <at> gnu.org>, 39823 <at> debbugs.gnu.org, rms <at> gnu.org
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Mon, 22 Jun 2020 11:07:25 -0400
Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:

> +  (let* (;; Honor the :generated-autoload-file keyword argument.
> +         (generated-autoload-file-arg (seq-position
> +                                       dirs ':generated-autoload-file))
> +         (generated-autoload-file (if generated-autoload-file-arg
> +                                      (elt dirs
> +                                           (1+ generated-autoload-file-arg))
> +                                    generated-autoload-file))
> +         ;; Cleanup dirs from such keyword argument, if any.
> +         (dirs (if generated-autoload-file-arg
> +                   (delete generated-autoload-file
> +                           (delete ':generated-autoload-file dirs))
> +                 dirs))

How about

    (let* ((generated-autoload-file
            (if (eq (car dirs) :generated-autoload-file)
                (progn (pop dirs) (pop dirs))
              generated-autoload-file))





Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Tue, 23 Jun 2020 20:43:02 GMT) Full text and rfc822 format available.

Message #59 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Noam Postavsky <npostavs <at> gmail.com>
Cc: Eli Zaretskii <eliz <at> gnu.org>, 39823 <at> debbugs.gnu.org, rms <at> gnu.org
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Tue, 23 Jun 2020 16:42:33 -0400
Hello Noam,

Noam Postavsky <npostavs <at> gmail.com> writes:

> Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:
>
>> +  (let* (;; Honor the :generated-autoload-file keyword argument.
>> +         (generated-autoload-file-arg (seq-position
>> +                                       dirs ':generated-autoload-file))
>> +         (generated-autoload-file (if generated-autoload-file-arg
>> +                                      (elt dirs
>> +                                           (1+ generated-autoload-file-arg))
>> +                                    generated-autoload-file))
>> +         ;; Cleanup dirs from such keyword argument, if any.
>> +         (dirs (if generated-autoload-file-arg
>> +                   (delete generated-autoload-file
>> +                           (delete ':generated-autoload-file dirs))
>> +                 dirs))
>
> How about
>
>     (let* ((generated-autoload-file
>             (if (eq (car dirs) :generated-autoload-file)
>                 (progn (pop dirs) (pop dirs))
>               generated-autoload-file))

That's much simpler, which is nice!  The only drawback is that it
requires the keyword argument to be used as the first argument when
provided (mine didn't have such restriction).

Provided the docstring reflects this, I'm OK with it :-).

Thank you,

Maxim




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Wed, 30 Sep 2020 15:24:01 GMT) Full text and rfc822 format available.

Message #62 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Lars Ingebrigtsen <larsi <at> gnus.org>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org, Eli Zaretskii <eliz <at> gnu.org>,
 Noam Postavsky <npostavs <at> gmail.com>, rms <at> gnu.org
Subject: Re: bug#39823: 26.3; update-directory-autoloads regression from
 Emacs 26 to Emacs 27
Date: Wed, 30 Sep 2020 17:23:21 +0200
Maxim Cournoyer <maxim.cournoyer <at> gmail.com> writes:

> I've given this idea a try with the attached patch.  For some reason I
> can't seem to simply eval the definition to test it (Debugger
> entered--Lisp error: (void-function byte-compile-info-string).  I'm
> rebuilding Emacs with the change right now.

[...]

> This provides a way to explicitly specify the GENERATED-AUTOLOAD-FILE
> file name to use for the generated autoload definitions, which is more
> convenient when using lexical scoping.  Previously, the user had to
> use dynamic scoping and bind the variable before calling
> `update-directory-autoloads'.  This was discussed in
> <https://bugs.gnu.org/39823>.

This seems like a good change -- I'm all for explicit function arguments
instead of binding variables around the call.

[...]

>  ;;;###autoload
> -(defun update-directory-autoloads (&rest dirs)
> +(cl-defun update-directory-autoloads (&rest dirs
> +                                            &key (generated-autoload-file
> +                                                  generated-autoload-file)
> +                                            &allow-other-keys)

But I'm not really enthusiastic about this complicated function
signature.  The doc string to the function already goes on way too long
to explain what the parameter means.

I think it would make more sense to introduce a new function, say,
`make-directory-autoloads' (because it doesn't so much update the
autoloads as create an autoload file, I think?) with a sane function
signature, and then deprecate update-directory-autoloads (which is only
used a handful of place in the Emacs tree).

Any comments?

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




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Thu, 01 Oct 2020 17:21:02 GMT) Full text and rfc822 format available.

Message #65 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Lars Ingebrigtsen <larsi <at> gnus.org>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, Noam Postavsky <npostavs <at> gmail.com>
Subject: Re: bug#39823: 26.3; update-directory-autoloads regression from
 Emacs 26 to Emacs 27
Date: Thu, 01 Oct 2020 19:20:01 +0200
Lars Ingebrigtsen <larsi <at> gnus.org> writes:

> I think it would make more sense to introduce a new function, say,
> `make-directory-autoloads' (because it doesn't so much update the
> autoloads as create an autoload file, I think?) with a sane function
> signature, and then deprecate update-directory-autoloads (which is only
> used a handful of place in the Emacs tree).

Man, that was a lot more invasive than I thought, because the
dynamically bound variable was used all over the place.

But with the following patch

(make-directory-autoloads "~/src/emacs/trunk/lisp/calc" "/tmp/autofoo.el")

works.  Emacs builds after applying the patch, and starts, and no tests
fail.

Does this look OK to everybody?

diff --git a/doc/lispref/loading.texi b/doc/lispref/loading.texi
index 6833af9c26..aa6ef307b1 100644
--- a/doc/lispref/loading.texi
+++ b/doc/lispref/loading.texi
@@ -577,7 +577,7 @@ Autoload
 define function @var{function-name}"}.
 
 @findex update-file-autoloads
-@findex update-directory-autoloads
+@findex make-directory-autoloads
 @cindex magic autoload comment
 @cindex autoload cookie
 @anchor{autoload cookie}
@@ -590,7 +590,7 @@ Autoload
 file generated by @code{update-file-autoloads} can be changed from the
 above defaults, see below.)
 Building Emacs loads @file{loaddefs.el} and thus calls @code{autoload}.
-@kbd{M-x update-directory-autoloads} is even more powerful; it updates
+@kbd{M-x make-directory-autoloads} is even more powerful; it updates
 autoloads for all files in the current directory.
 
   The same magic comment can copy any kind of form into
diff --git a/lisp/emacs-lisp/autoload.el b/lisp/emacs-lisp/autoload.el
index 0bcd7b7dcd..fe6a72c0a1 100644
--- a/lisp/emacs-lisp/autoload.el
+++ b/lisp/emacs-lisp/autoload.el
@@ -254,13 +254,12 @@ make-autoload
 ;; the doc-string in FORM.
 ;; Those properties are now set in lisp-mode.el.
 
-(defun autoload-find-generated-file ()
+(defun autoload-find-generated-file (file)
   "Visit the autoload file for the current buffer, and return its buffer."
   (let ((enable-local-variables :safe)
         (enable-local-eval nil)
         (find-file-hook nil)
-        (delay-mode-hooks t)
-        (file (autoload-generated-file)))
+        (delay-mode-hooks t))
     ;; We used to use `raw-text' to read this file, but this causes
     ;; problems when the file contains non-ASCII characters.
     (with-current-buffer (find-file-noselect
@@ -268,11 +267,13 @@ autoload-find-generated-file
       (if (zerop (buffer-size)) (insert (autoload-rubric file nil t)))
       (current-buffer))))
 
-(defun autoload-generated-file ()
+(defun autoload-generated-file (outfile)
   "Return `generated-autoload-file' as an absolute name.
 If local to the current buffer, expand using the default directory;
 otherwise, using `source-directory'/lisp."
-  (expand-file-name generated-autoload-file
+  (expand-file-name (if (local-variable-p 'generated-autoload-file)
+                        generated-autoload-file
+                      outfile)
                     ;; File-local settings of generated-autoload-file should
                     ;; be interpreted relative to the file's location,
                     ;; of course.
@@ -454,13 +455,12 @@ autoload-find-file
 (defvar no-update-autoloads nil
   "File local variable to prevent scanning this file for autoload cookies.")
 
-(defun autoload-file-load-name (file)
+(defun autoload-file-load-name (file outfile)
   "Compute the name that will be used to load FILE."
   ;; OUTFILE should be the name of the global loaddefs.el file, which
   ;; is expected to be at the root directory of the files we're
   ;; scanning for autoloads and will be in the `load-path'.
-  (let* ((outfile (default-value 'generated-autoload-file))
-         (name (file-relative-name file (file-name-directory outfile)))
+  (let* ((name (file-relative-name file (file-name-directory outfile)))
          (names '())
          (dir (file-name-directory outfile)))
     ;; If `name' has directory components, only keep the
@@ -608,7 +608,7 @@ autoload--make-defs-autoload
         `(register-definition-prefixes ,file ',(sort (delq nil strings)
 						     'string<))))))
 
-(defun autoload--setup-output (otherbuf outbuf absfile load-name)
+(defun autoload--setup-output (otherbuf outbuf absfile load-name output-file)
   (let ((outbuf
          (or (if otherbuf
                  ;; A file-local setting of
@@ -616,7 +616,7 @@ autoload--setup-output
                  ;; should ignore OUTBUF.
                  nil
                outbuf)
-             (autoload-find-destination absfile load-name)
+             (autoload-find-destination absfile load-name output-file)
              ;; The file has autoload cookies, but they're
              ;; already up-to-date. If OUTFILE is nil, the
              ;; entries are in the expected OUTBUF,
@@ -717,7 +717,7 @@ autoload-generate-file-autoloads
                   (setq load-name
                         (if (stringp generated-autoload-load-name)
                             generated-autoload-load-name
-                          (autoload-file-load-name absfile)))
+                          (autoload-file-load-name absfile outfile)))
                   ;; FIXME? Comparing file-names for equality with just equal
                   ;; is fragile, eg if one has an automounter prefix and one
                   ;; does not, but both refer to the same physical file.
@@ -725,8 +725,10 @@ autoload-generate-file-autoloads
                              (not
                               (if (memq system-type '(ms-dos windows-nt))
                                   (equal (downcase outfile)
-                                         (downcase (autoload-generated-file)))
-                                (equal outfile (autoload-generated-file)))))
+                                         (downcase (autoload-generated-file
+                                                    outfile)))
+                                (equal outfile (autoload-generated-file
+                                                outfile)))))
                     (setq otherbuf t))
                   (save-excursion
                     (save-restriction
@@ -740,7 +742,8 @@ autoload-generate-file-autoloads
                                                  (file-name-sans-extension
                                                   (file-name-nondirectory file))))
                                (setq output-start (autoload--setup-output
-                                                   otherbuf outbuf absfile load-name))
+                                                   otherbuf outbuf absfile
+                                                   load-name outfile))
                                (let ((standard-output (marker-buffer output-start))
                                      (print-quoted t))
                                  (princ `(push (purecopy
@@ -758,7 +761,8 @@ autoload-generate-file-autoloads
                             ;; If not done yet, figure out where to insert this text.
                             (unless output-start
                               (setq output-start (autoload--setup-output
-                                                  otherbuf outbuf absfile load-name)))
+                                                  otherbuf outbuf absfile
+                                                  load-name outfile)))
                             (autoload--print-cookie-text output-start load-name file))
                            ((= (following-char) ?\;)
                             ;; Don't read the comment.
@@ -789,7 +793,7 @@ autoload-generate-file-autoloads
                ((not otherbuf)
                 (unless output-start
                   (setq output-start (autoload--setup-output
-                                      nil outbuf absfile load-name)))
+                                      nil outbuf absfile load-name outfile)))
                 (let ((autoload-print-form-outbuf
                        (marker-buffer output-start)))
                   (autoload-print-form form)))
@@ -801,9 +805,8 @@ autoload-generate-file-autoloads
                         ;; then passing otherbuf=nil is enough, but if
                         ;; outbuf is nil, that won't cut it, so we
                         ;; locally bind generated-autoload-file.
-                        (let ((generated-autoload-file
-                               (default-value 'generated-autoload-file)))
-                          (autoload--setup-output nil outbuf absfile load-name)))
+                        (autoload--setup-output nil outbuf absfile load-name
+                                                outfile))
                        (autoload-print-form-outbuf
                         (marker-buffer other-output-start)))
                   (autoload-print-form form)
@@ -937,7 +940,7 @@ update-file-autoloads
           (message "Autoload section for %s is up to date." file)))
     (if no-autoloads file)))
 
-(defun autoload-find-destination (file load-name)
+(defun autoload-find-destination (file load-name output-file)
   "Find the destination point of the current buffer's autoloads.
 FILE is the file name of the current buffer.
 LOAD-NAME is the name as it appears in the output.
@@ -947,12 +950,12 @@ autoload-find-destination
   (catch 'up-to-date
     (let* ((buf (current-buffer))
            (existing-buffer (if buffer-file-name buf))
-           (output-file (autoload-generated-file))
+           (output-file (autoload-generated-file output-file))
            (output-time (if (file-exists-p output-file)
                             (file-attribute-modification-time
 			     (file-attributes output-file))))
            (found nil))
-      (with-current-buffer (autoload-find-generated-file)
+      (with-current-buffer (autoload-find-generated-file output-file)
         ;; This is to make generated-autoload-file have Unix EOLs, so
         ;; that it is portable to all platforms.
         (or (eq 0 (coding-system-eol-type buffer-file-coding-system))
@@ -1038,7 +1041,20 @@ update-directory-autoloads
 use the existing value of `generated-autoload-file'.  If any Lisp
 file binds `generated-autoload-file' as a file-local variable,
 write its autoloads into the specified file instead."
-  (interactive "DUpdate autoloads from directory: ")
+  (declare (obsolete make-directory-autoloads "28.1"))
+  (make-directory-autoloads dirs generated-autoload-file))
+
+;;;###autoload
+(defun make-directory-autoloads (dir output-file)
+  "Update autoload definitions for Lisp files in the directories DIRS.
+DIR can be either a single directory or a list of
+directories.  (The latter usage is discouraged.)
+
+The autoloads will be written to OUTPUT-FILE.
+
+The function does NOT recursively descend into subdirectories of the
+directory or directories specified."
+  (interactive "DUpdate autoloads from directory: \nfWrite to file: ")
   (let* ((files-re (let ((tmp nil))
 		     (dolist (suf (get-load-suffixes))
                        ;; We don't use module-file-suffix below because
@@ -1049,10 +1065,10 @@ update-directory-autoloads
                          (push suf tmp)))
                      (concat "\\`[^=.].*" (regexp-opt tmp t) "\\'")))
 	 (files (apply #'nconc
-		       (mapcar (lambda (dir)
-				 (directory-files (expand-file-name dir)
-						  t files-re))
-			       dirs)))
+		       (mapcar (lambda (d)
+				 (directory-files (expand-file-name d)
+                                                  t files-re))
+			       (if (consp dir) dir (list dir)))))
          (done ())                      ;Files processed; to remove duplicates.
          (changed nil)                  ;Non-nil if some change occurred.
 	 (last-time)
@@ -1060,16 +1076,12 @@ update-directory-autoloads
          ;; files because of file-local autoload-generated-file settings.
 	 (no-autoloads nil)
          (autoload-modified-buffers nil)
-	 (generated-autoload-file
-	  (if (called-interactively-p 'interactive)
-	      (read-file-name "Write autoload definitions to file: ")
-	    generated-autoload-file))
 	 (output-time
-	  (if (file-exists-p generated-autoload-file)
-	      (file-attribute-modification-time
-	       (file-attributes generated-autoload-file)))))
+	  (and (file-exists-p output-file)
+	       (file-attribute-modification-time
+                (file-attributes output-file)))))
 
-    (with-current-buffer (autoload-find-generated-file)
+    (with-current-buffer (autoload-find-generated-file output-file)
       (save-excursion
 	;; Canonicalize file names and remove the autoload file itself.
 	(setq files (delete (file-relative-name buffer-file-name)
@@ -1126,8 +1138,7 @@ update-directory-autoloads
             (progress (make-progress-reporter
                        (byte-compile-info
                         (concat "Scraping files for "
-                                (file-relative-name
-                                 generated-autoload-file)))
+                                (file-relative-name output-file)))
                        0 (length files) nil 10))
             (file-count 0)
             file-time)
@@ -1190,6 +1201,8 @@ batch-update-autoloads
   ;; Exclude those files that are preloaded on ALL platforms.
   ;; These are the ones in loadup.el where "(load" is at the start
   ;; of the line (crude, but it works).
+  (message "FOOOO")
+  (setq debug-on-error t)
   (unless autoload-excludes
     (let ((default-directory (file-name-directory generated-autoload-file))
 	  file)
@@ -1205,7 +1218,7 @@ batch-update-autoloads
   (let ((args command-line-args-left))
     (batch-update-autoloads--summary args)
     (setq command-line-args-left nil)
-    (apply #'update-directory-autoloads args)))
+    (make-directory-autoloads args generated-autoload-file)))
 
 (provide 'autoload)
 
diff --git a/lisp/emacs-lisp/package.el b/lisp/emacs-lisp/package.el
index a173fc060a..e77077fc52 100644
--- a/lisp/emacs-lisp/package.el
+++ b/lisp/emacs-lisp/package.el
@@ -1013,7 +1013,6 @@ package-autoload-ensure-default-file
     (write-region (autoload-rubric file "package" nil) nil file nil 'silent))
   file)
 
-(defvar generated-autoload-file)
 (defvar autoload-timestamps)
 (defvar version-control)
 
@@ -1021,14 +1020,14 @@ package-generate-autoloads
   "Generate autoloads in PKG-DIR for package named NAME."
   (let* ((auto-name (format "%s-autoloads.el" name))
          ;;(ignore-name (concat name "-pkg.el"))
-         (generated-autoload-file (expand-file-name auto-name pkg-dir))
+         (output-file (expand-file-name auto-name pkg-dir))
          ;; We don't need 'em, and this makes the output reproducible.
          (autoload-timestamps nil)
          (backup-inhibited t)
          (version-control 'never))
-    (package-autoload-ensure-default-file generated-autoload-file)
-    (update-directory-autoloads pkg-dir)
-    (let ((buf (find-buffer-visiting generated-autoload-file)))
+    (package-autoload-ensure-default-file output-file)
+    (make-directory-autoloads pkg-dir output-file)
+    (let ((buf (find-buffer-visiting output-file)))
       (when buf (kill-buffer buf)))
     auto-name))
 
diff --git a/test/lisp/vc/vc-bzr-tests.el b/test/lisp/vc/vc-bzr-tests.el
index b68a694512..408d6e8e23 100644
--- a/test/lisp/vc/vc-bzr-tests.el
+++ b/test/lisp/vc/vc-bzr-tests.el
@@ -131,7 +131,6 @@ vc-bzr-test-faulty-bzr-autoloads
                  (make-directory bzrdir)
                  (expand-file-name "foo.el" bzrdir)))
          (default-directory (file-name-as-directory bzrdir))
-         (generated-autoload-file (expand-file-name "loaddefs.el" bzrdir))
          (process-environment (cons (format "HOME=%s" homedir)
                                     process-environment)))
     (unwind-protect
@@ -148,7 +147,9 @@ vc-bzr-test-faulty-bzr-autoloads
           ;; causes bzr status to fail.  This simulates a broken bzr
           ;; installation.
           (delete-file ".bzr/checkout/dirstate")
-          (should (progn (update-directory-autoloads default-directory)
+          (should (progn (make-directory-autoloads
+                          default-directory
+                          (expand-file-name "loaddefs.el" bzrdir))
                          t)))
       (delete-directory homedir t))))
 


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




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Thu, 01 Oct 2020 17:55:01 GMT) Full text and rfc822 format available.

Message #68 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Lars Ingebrigtsen <larsi <at> gnus.org>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, maxim.cournoyer <at> gmail.com,
 npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3;
 update-directory-autoloads regression from Emacs 26 to Emacs 27
Date: Thu, 01 Oct 2020 20:54:07 +0300
> From: Lars Ingebrigtsen <larsi <at> gnus.org>
> Date: Thu, 01 Oct 2020 19:20:01 +0200
> Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, Noam Postavsky <npostavs <at> gmail.com>
> 
> Lars Ingebrigtsen <larsi <at> gnus.org> writes:
> 
> > I think it would make more sense to introduce a new function, say,
> > `make-directory-autoloads' (because it doesn't so much update the
> > autoloads as create an autoload file, I think?) with a sane function
> > signature, and then deprecate update-directory-autoloads (which is only
> > used a handful of place in the Emacs tree).
> 
> Man, that was a lot more invasive than I thought, because the
> dynamically bound variable was used all over the place.
> 
> But with the following patch
> 
> (make-directory-autoloads "~/src/emacs/trunk/lisp/calc" "/tmp/autofoo.el")
> 
> works.  Emacs builds after applying the patch, and starts, and no tests
> fail.
> 
> Does this look OK to everybody?

Before we go and make all these changes, how about considering the
solution I proposed up-thread, i.e. leave this alone?  I mean, is it
really worth all this?  It's an obscure use case with a very easy
solution/workaround.

But feel free to ignore me.




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Thu, 01 Oct 2020 18:07:02 GMT) Full text and rfc822 format available.

Message #71 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Lars Ingebrigtsen <larsi <at> gnus.org>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, maxim.cournoyer <at> gmail.com,
 npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3; update-directory-autoloads regression from
 Emacs 26 to Emacs 27
Date: Thu, 01 Oct 2020 20:06:14 +0200
Eli Zaretskii <eliz <at> gnu.org> writes:

> Before we go and make all these changes, how about considering the
> solution I proposed up-thread, i.e. leave this alone?  I mean, is it
> really worth all this?  It's an obscure use case with a very easy
> solution/workaround.

I just feel that the logic in these functions is kinda obscure, what
with the mix of dynamically-bound variables, function parameters and
local variables.  The patch removes the former from the mix, making them
easier to reason about (and read).

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




Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Fri, 02 Oct 2020 02:37:02 GMT) Full text and rfc822 format available.

Message #74 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Lars Ingebrigtsen <larsi <at> gnus.org>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 39823 <at> debbugs.gnu.org, rms <at> gnu.org, maxim.cournoyer <at> gmail.com,
 npostavs <at> gmail.com
Subject: Re: bug#39823: 26.3; update-directory-autoloads regression from
 Emacs 26 to Emacs 27
Date: Fri, 02 Oct 2020 04:36:08 +0200
Lars Ingebrigtsen <larsi <at> gnus.org> writes:

> I just feel that the logic in these functions is kinda obscure, what
> with the mix of dynamically-bound variables, function parameters and
> local variables.  The patch removes the former from the mix, making them
> easier to reason about (and read).

I went ahead and pushed this change (which also fixes bug#38694) even if
it was more invasive than I'd have preferred.  I think the new interface
is better, and anyway: Sunk cost fallacy.

Let me know if I broke anything.

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




Added tag(s) fixed. Request was from Lars Ingebrigtsen <larsi <at> gnus.org> to control <at> debbugs.gnu.org. (Fri, 02 Oct 2020 02:37:02 GMT) Full text and rfc822 format available.

bug marked as fixed in version 28.1, send any further explanations to 39823 <at> debbugs.gnu.org and Maxim Cournoyer <maxim.cournoyer <at> gmail.com> Request was from Lars Ingebrigtsen <larsi <at> gnus.org> to control <at> debbugs.gnu.org. (Fri, 02 Oct 2020 02:37:03 GMT) Full text and rfc822 format available.

Information forwarded to bug-gnu-emacs <at> gnu.org:
bug#39823; Package emacs. (Fri, 02 Oct 2020 17:53:02 GMT) Full text and rfc822 format available.

Message #81 received at 39823 <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Lars Ingebrigtsen <larsi <at> gnus.org>
Cc: Eli Zaretskii <eliz <at> gnu.org>, npostavs <at> gmail.com, 39823 <at> debbugs.gnu.org,
 rms <at> gnu.org
Subject: Re: bug#39823: 26.3; update-directory-autoloads regression from
 Emacs 26 to Emacs 27
Date: Fri, 02 Oct 2020 13:54:11 -0400
Hello,

Lars Ingebrigtsen <larsi <at> gnus.org> writes:

> Lars Ingebrigtsen <larsi <at> gnus.org> writes:
>
>> I just feel that the logic in these functions is kinda obscure, what
>> with the mix of dynamically-bound variables, function parameters and
>> local variables.  The patch removes the former from the mix, making them
>> easier to reason about (and read).
>
> I went ahead and pushed this change (which also fixes bug#38694) even if
> it was more invasive than I'd have preferred.  I think the new interface
> is better, and anyway: Sunk cost fallacy.

Thanks for this work!  I look forward to use the improved, more
functional `make-directory-autoloads' procedure.

Maxim




bug archived. Request was from Debbugs Internal Request <help-debbugs <at> gnu.org> to internal_control <at> debbugs.gnu.org. (Sat, 31 Oct 2020 11:24:05 GMT) Full text and rfc822 format available.

This bug report was last modified 3 years and 177 days ago.

Previous Next


GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.