* lisp/emacs-lisp/loaddefs-gen.el (loaddefs-generate): Always
generate an output file if we have EXTRA-DATA. This fixes package
installation of packages with no ;;;###autoload forms.
Commit 30ec4a7347 is incorrect -- the
length computation logic uses a simple nilness test, not `numberp'.
The `numberp' case is just an optimization if `len' is a literal
number; it does not affect the behavior.
Revert that commit, add some comments to help future readers avoid the
same mistake, and update the pack logic to use the same optimization
as the length computation for consistency.
* lisp/emacs-lisp/bytecomp.el (byte-compile--reify-function):
Don't transpose doc string and interactive spec, which must come
in this order.
* test/lisp/emacs-lisp/bytecomp-tests.el
(bytecomp-function-attributes): New test.
* lisp/emacs-lisp/loaddefs-gen.el (loaddefs-generate--rubric): Fix
the name that's recorded in the file -- this is only used by
loaddefs-gen now, I think.
* lisp/emacs-lisp/find-func.el (find-function-advised-original): Look
at the actual function definition rather than
`advice--symbol-function` which sometimes returns only the advice.
* test/lisp/emacs-lisp/find-func-tests.el
(find-func-tests--find-library-verbose): Add test.
`cl-typep` has used a heuristic that if there's a `<foo>-p` function,
then <foo> can be used as a type. This made sense in the past where
most types were not officially declared to be (cl-)types, but nowadays
this just encourages abuses such as using `cl-typecase` with
"types" like `fbound`. It's also a problem for EIEIO objects, where
for historical reasons `<foo>-p` tests if the object is of type
exactly `<foo>` whereas (cl-typep OBJ <foo>) should instead test
if OBJ is a *subtype* of `<foo>`.
So we change `cl-typep` to emit a warning whenever this "-p" heuristic
is used, to discourage abuses, encourage the use of explicit
`cl-deftype` declarations, and try and detect some misuses of
`<foo>-p` for EIEIO objects.
* lisp/emacs-lisp/eieio.el (defclass): Define as type not only at
run-time but also for the current compilation unit.
* lisp/emacs-lisp/eieio-core.el (class, eieio-object): Define as types.
* lisp/emacs-lisp/cl-preloaded.el (cl-struct-define): Don't abuse the
"-p" heuristic.
* lisp/emacs-lisp/cl-macs.el (cl-deftype-satisfies):
Add entries for frames, windows, markers, and overlays.
(cl-typep): Emit a warning when using a predicate that is not known to
correspond to a type.
* lisp/files.el (file-relative-name): Fix error that can trigger if
there's an(other) error between loading `files.el` and loading
`minibuffer.el`.
* lisp/emacs-lisp/loaddefs-gen.el (generated-autoload-file)
(generated-autoload-load-name): Move the remaining two autoload.el
variables used by loaddefs-gen.el from autoload.el.
* lisp/url/url-cookie.el (url-cookie-write-file):
* lisp/international/titdic-cnv.el (tit-process-header):
* lisp/international/ja-dic-cnv.el (skkdic-convert):
* lisp/international/emoji.el (emoji--generate-file):
* lisp/emacs-lisp/loaddefs-gen.el (loaddefs-generate--rubric)
* admin/unidata/unidata-gen.el (unidata-gen-file)
(unidata-gen-charprop): Use the new functions.
* lisp/emacs-lisp/generate-file.el: New file to provide
convenience functions for generated files. It's not always
trivial to know which parts of the trailer that has to be
obfuscated to avoid not getting byte-compiled etc, and some parts
of the headers/trailers are usually forgotten when hand-coding
these.
* lisp/emacs-lisp/shortdoc.el (string): Each example section is
supposed to contain only examples of using the function in
question (as policed by the FAILED shortdoc-examples test).
* lisp/emacs-lisp/checkdoc.el (checkdoc-file-comments-engine): Use
the more general lisp-mode-autoload-regexp instead of
generate-autoload-cookie (i.e., also match ;;;###tramp-autoload).
* lisp/Makefile.in ($(lisp)/loaddefs.el): Use the new function.
* lisp/emacs-lisp/loaddefs-gen.el (loaddefs-generate): Pass in
whether to inhibit a partial build (to make the code more general).
(loaddefs-generate--emacs-batch): Add a new function specially for
the Emacs build that has the special rules needed. (This also
fixes out-of-tree builds.)
loaddefs-generate-batch can be used in general for packages etc.
(loaddefs-generate-batch): Remove the special code for Emacs builds.
* lisp/emacs-lisp/bindat.el (strz): Fix (wrong-type-argument
number-or-marker-p nil) error when unpacking a strz with
unspecified (variable) length.
* test/lisp/emacs-lisp/bindat-tests.el (strz): Mark test as passing.
* lisp/emacs-lisp/bindat.el (strz): Include null terminator when
computing packed string length.
* test/lisp/emacs-lisp/bindat-tests.el (strz): Mark tests as passing.
* lisp/emacs-lisp/loaddefs-gen.el (loaddefs-generate): Do a
complete build more often to avoid problems with going from old
loaddefs.el files to new ones.
* lisp/emacs-lisp/loaddefs-gen.el (loaddefs-generate): Only update
the loaddefs section for the files that have changed. This makes
"git pull" (when few/no Lisp files have changed much faster).
* doc/lispref/loading.texi (Autoload, Autoload by Prefix): Refer
to loaddefs-generate instead of update-file-autoloads.
* lisp/Makefile.in (LOADDEFS): Remove, because all the loaddefs
files are created in one go now.
(COMPILE_FIRST): Add loaddefs-gen/radix-tree, and drop autoload.
($(lisp)/loaddefs.el): Use loaddefs-gen.
(MH_E_DIR, $(TRAMP_DIR)/tramp-loaddefs.el)
($(MH_E_DIR)/mh-loaddefs.el, $(CAL_DIR)/cal-loaddefs.el)
($(CAL_DIR)/diary-loaddefs.el, $(CAL_DIR)/hol-loaddefs.el): Remove.
* lisp/generic-x.el: Inhibit computing prefixes, because the
namespace here is all wonky.
* lisp/w32-fns.el (w32-batch-update-autoloads): Removed -- unused
function.
* lisp/calendar/holidays.el ("holiday-loaddefs"): Renamed from
hol-loaddefs to have a more regular name.
* lisp/cedet/ede/proj-elisp.el (ede-emacs-cedet-autogen-compiler):
Refer to loaddefs-gen instead of autoload.
* lisp/emacs-lisp/autoload.el (make-autoload, autoload-rubric)
(autoload-insert-section-header): Made into aliases of
loaddefs-gen functions.
(autoload--make-defs-autoload): Ditto.
(autoload-ignored-definitions, autoload-compute-prefixes): Moved
to loaddefs-gen.
* lisp/emacs-lisp/lisp-mode.el (lisp-mode-autoload-regexp): New
constant.
(lisp-fdefs, lisp-mode-variables, lisp-outline-level): Use it to
recognize all ;;;###autoload forms.
* lisp/emacs-lisp/loaddefs-gen.el: New file.
* lisp/emacs-lisp/package.el: Use loaddefs-generate instead of
make-directory-autoloads.
* test/lisp/vc/vc-bzr-tests.el (vc-bzr-test-faulty-bzr-autoloads):
Use loaddefs instead of autoloads.
* lisp/emacs-lisp/bytecomp.el (byte-compile--first-symbol-with-pos):
Traverse record literals as well as vectors. Either is rather
pointless but there were some strong feelings about it.
The bytecode interpreter can't directly call special forms, so
the byte-compiler usually converts special forms into some sequence of
byte codes (basically, providing a duplicate definition of the special
form). There are still two exceptions to this: `defconst` and `defvar`,
where the compiler instead generates a convoluted chunk of code like:
(funcall '(lambda (x) (defvar <sym> x <doc>)) <value>)
where the quote makes sure we keep the function non-compiled, so as
to end up running the special form at run time.
Get rid of this workaround by introducing `defvar-1` and `defconst-1`
which provide a *functional* interface to the functionality of the
corresponding special form.
* src/eval.c (defvar, Fdefvar_1, Fdefconst_1): New functions, extracted from
`Fdefvar` and `Fdefconst`.
(Fdefvar, Fdefconst): Use them.
(syms_of_eval): `defsubr` the new functions.
* lisp/emacs-lisp/bytecomp.el (byte-compile-tmp-var): Delete const.
(byte-compile-defvar): Simplify using the new functions.
* doc/lispref/variables.texi (Defining Variables): Adjust the doc of
`defvar` to reflect the actual semantics implemented.
* lisp/emacs-lisp/bytecomp.el (byte-compile--first-symbol-with-pos)
(byte-compile--warning-source-offset):
Remove recursion for cdr-traversal of lists, and optimise (bug#55414).
