Improve the handling of `or` patterns where not all sub-patterns bind the
same set of variables. This used to be "unsupported" and behaved in
somewhat unpredictable ways.
(pcase--expand): Rewrite.
(pcase-codegen): Delete.
* doc/lispref/control.texi (pcase Macro): Adjust accordingly.
Also remove the warning about "at least two" sub patterns.
These work fine, AFAICT, and if not we should fix it.
* test/lisp/emacs-lisp/pcase-tests.el (pcase-tests-or-vars): New test.
Revert commit a218c98615, but in order
to avoid the spurious warnings that this commit tried to squash,
keep track of the vars used during the match so as to add
corresponding annotations to explicitly silence the spurious warnings.
To do this, we change the VARS used in `pcase-u` (and throughout
the pcase code): they used to hold elements of the form (NAME . VAL)
and now they hold elements of the form (NAME VAL . USED).
(pcase--expand): Bind all vars instead of only those found via fgrep.
(pcase-codegen): Silence "unused var" warnings for those vars that have
already been referenced during the match itself.
(pcase--funcall, pcase--eval): Record the vars that are used.
(pcase--u1): Record the vars that are used via non-linear patterns.
* lisp/textmodes/mhtml-mode.el (mhtml-forward):
* lisp/vc/diff-mode.el (diff-goto-source): Silence newly
discovered warnings.
* test/lisp/emacs-lisp/pcase-tests.el (pcase-tests-bug46786): New test.
The pcase 'rx' pattern would in some cases allow the match data to be
clobbered before it is read. For example:
(pcase "PQR"
((and (rx (let a nonl)) (rx ?z)) (list 'one a))
((rx (let b ?Q)) (list 'two b)))
The above returned (two "P") instead of the correct (two "Q").
This occurred because the calls to string-match and match-string were
presented as separate patterns to pcase, which would interleave them
with other patterns.
As a remedy, combine string matching and match-data extraction into a
single pcase pattern. This introduces a slight inefficiency for two
or more submatches as they are grouped into a list structure which
then has to be destructured.
Found by Stefan Monnier. See discussion at
https://lists.gnu.org/archive/html/emacs-devel/2021-02/msg02010.html
* lisp/emacs-lisp/rx.el (rx--reduce-right): New helper.
(rx [pcase macro]): Combine string-match and match-string calls into a
single pcase pattern.
* test/lisp/emacs-lisp/rx-tests.el (rx-pcase): Add test cases.
Two unrelated bugs: A missing type check caused an error in rx
patterns for non-string match targets, and rx patterns did not work at
all in pcase-let or pcase-let*.
Second bug reported by Basil Contovounesios and Ag Ibragimov; fixes
proposed by Stefan Monnier. Discussion and explanation in thread at
https://lists.gnu.org/archive/html/emacs-devel/2021-02/msg01924.html
* lisp/emacs-lisp/rx.el (rx): Add (pred stringp) to avoid type errors,
and replace the `pred` clause for the actual match with something that
works with pcase-let(*) without being optimised away.
* test/lisp/emacs-lisp/rx-tests.el (rx-pcase): Add test cases.
Add use of `macroexp-file-name` from a macro called from within
a function, which works thanks to eager-macroexpansion (so the macro
is expanded which the file is being loaded rather than only later
when the function is called).
* test/lisp/emacs-lisp/macroexp-resources/m1.el
(macroexp--m1-tests-file-name): New function.
(edebug-tests-trivial-backquote): Adjust to the way `eval-defun`
outputs its result.
(edebug-tests-cl-macrolet): Adjust to the fact that now macro expansion
takes place during the `eval-defun` even when Edebugging.
* lisp/progmodes/cperl-mode.el (cperl--package-regexp):
Avoid double repetition; cperl--ws-or-comment-regexp is already
repeated with 1+.
* test/lisp/textmodes/dns-mode-tests.el
(dns-mode-tests-dns-mode-soa-increment-serial): Escape literal '$'.
* test/lisp/emacs-lisp/rx-tests.el (rx-regexp): Modify test to not
trigger a linting warning while retaining its testing power.
The argument of the rx `regexp` form is assumed to evaluate to a valid
regexp, but certain kinds of deprecated but still accepted usage were
not handled correctly, such as unescaped literal (special) characters:
(rx "a" (regexp "*")) => "a*" which is wrong.
Handle these cases; there is no extra trouble.
* lisp/emacs-lisp/rx.el (rx--translate-regexp): Force bracketing
of single special characters.
* test/lisp/emacs-lisp/rx-tests.el (rx-regexp): Add test case.
It's basically an alias for `quote`, but it offers the advantage of
providing Edebug support and opens the possibility of compiling
the bindat spec to ELisp code.
* doc/lispref/processes.texi (Bindat Spec): Document `bindat-spec`.
(Bindat Functions): Tweak a few things to adjust to the state of the code.
* test/lisp/emacs-lisp/bindat-tests.el: Use it.
* test/lisp/emacs-lisp/edebug-tests.el (edebug-tests--read): New function.
(edebug-tests--&rest-behavior): New test.
This allows the use of (declare (debug ...)) in the lexical macros
defined with `cl-macrolet`. It also fixes the names used by Edebug
for the methods of `cl-generic` so it doesn't need to use gensym
and so they don't include the formal arg names any more.
* lisp/emacs-lisp/edebug.el (edebug--match-&-spec-op):
Rename from `edebug--handle-&-spec-op`.
(edebug--match-&-spec-op <&interpose>): Rename from `&lookup` and
generalize so it can let-bind dynamic variables around the rest of the parse.
(edebug-lexical-macro-ctx): Rename from `edebug--cl-macrolet-defs` and
make it into an alist.
(edebug-list-form-args): Use the specs from `edebug-lexical-macro-ctx`
when available.
(edebug--current-cl-macrolet-defs): Delete var.
(edebug-match-cl-macrolet-expr, edebug-match-cl-macrolet-name)
(edebug-match-cl-macrolet-body): Delete functions.
(def-declarations): Use new `&interpose`.
(edebug--match-declare-arg): Rename from `edebug--get-declare-spec` and
adjust to new calling convention.
* lisp/subr.el (def-edebug-elem-spec): Fix docstring.
(eval-after-load): Use `declare`.
* lisp/emacs-lisp/cl-generic.el: Fix Edebug names so we don't need
gensym any more and we only include the specializers but not the formal
arg names.
(cl--generic-edebug-name): New var.
(cl--generic-edebug-remember-name, cl--generic-edebug-make-name): New funs.
(cl-defgeneric, cl-defmethod): Use them.
* lisp/emacs-lisp/cl-macs.el: Add support for `debug` declarations in
`cl-macrolet`.
(cl-declarations-or-string):
Fix use of `lambda-doc` and allow use of `declare`.
(edebug-lexical-macro-ctx): Declare var.
(cl--edebug-macrolet-interposer): New function.
(cl-macrolet): Use it to pass the right `lexical-macro-ctx` to the body.
* lisp/emacs-lisp/pcase.el (pcase-PAT): Use new `&interpose`.
(pcase--edebug-match-pat-args): Rename from `pcase--get-edebug-spec` and
adjust to new calling convention.
* test/lisp/emacs-lisp/cl-generic-tests.el (cl-defgeneric/edebug/method):
Adjust to the new names.
* test/lisp/emacs-lisp/edebug-tests.el (edebug-cl-defmethod-qualifier)
(edebug-tests-cl-flet): Adjust to the new names.
* doc/lispref/edebug.texi (Specification List): Document &interpose.
Arrange for declarations to be able to specify their own specs via
the `edebug-declaration-spec` property.
* lisp/emacs-lisp/edebug.el: (edebug--get-declare-spec): New function.
(def-declarations): New spec element.
(defun, defmacro): Use it in their spec.
* lisp/emacs-lisp/gv.el (gv-expander, gv-setter):
Set `edebug-declaration-spec`.
* test/lisp/emacs-lisp/edebug-tests.el (edebug-tests-gv-expander): New test.
* test/lisp/emacs-lisp/edebug-resources/edebug-test-code.el
(edebug-test-code-use-gv-expander): New test case.
The `edebug-form-spec` symbol property was used both to map forms's
head symbol to the corresponding spec, and to map spec element names
to their expansion.
This lead to name conflicts which break instrumentation of examples such as
(cl-flet ((gate (x) x)) (gate 4))
because of the Edebug spec element `gate`.
So introduce a new symbol property `edebug-elem-spec`.
* lisp/subr.el (def-edebug-elem-spec): New function.
* lisp/emacs-lisp/edebug.el (edebug--get-elem-spec): New function.
(edebug-match-symbol): Use it.
(Core Edebug elems): Put them on `edebug-elem-spec` instead of
`edebug-form-spec`.
(ELisp special forms): Set their `edebug-form-spec` via dolist.
(Other non-core Edebug elems): Use `def-edebug-elem-spec`.
(edebug-\`): Use `declare`.
* lisp/emacs-lisp/pcase.el (pcase-PAT, pcase-FUN, pcase-QPAT):
* lisp/skeleton.el (skeleton-edebug-spec):
* lisp/emacs-lisp/cl-macs.el: Use `def-edebug-elem-spec`.
* test/lisp/emacs-lisp/edebug-tests.el
(edebug-tests--conflicting-internal-names): New test.
* test/lisp/emacs-lisp/edebug-resources/edebug-test-code.el
(edebug-test-code-cl-flet1): New test case.
* doc/lispref/edebug.texi (Specification List): Add `def-edebug-elem-spec`.
(Specification Examples): Use it.
* doc/lispref/loading.texi (Hooks for Loading): Avoid the use of
`def-edebug-spec` in example (better use `debug` declaration).
This bug was introduced by the lexical variable constant propagation
mechanism. It was discovered by Michael Heerdegen.
* lisp/emacs-lisp/byte-opt.el (byte-optimize-let-form)
(byte-optimize-body): Let the effects of a local defvar declaration be
scoped by let and let*, not any arbitrary Lisp expression body (such
as progn).
* test/lisp/emacs-lisp/bytecomp-tests.el (bytecomp-tests--get-vars)
(bytecomp-local-defvar): New test.
Lexical variables bound to a constant value (symbol, number or string)
are substituted at their point of use and the variable then eliminated
if possible. Example:
(let ((x (+ 2 3))) (f x)) => (f 5)
This reduces code size, eliminates stack operations, and enables
further optimisations. The implementation is conservative, and is
strongly curtailed by the presence of variable mutation, conditions
and loops.
* lisp/emacs-lisp/byte-opt.el
(byte-optimize-enable-variable-constprop)
(byte-optimize-warn-eliminated-variable): New constants.
(byte-optimize--lexvars, byte-optimize--vars-outside-condition)
(byte-optimize--vars-outside-loop, byte-optimize--dynamic-vars):
New dynamic variables.
(byte-optimize--substitutable-p, byte-optimize-let-form):
New functions.
(byte-optimize-form-code-walker): Adapt clauses for variable
constprop, and add clauses for 'setq' and 'defvar'.
* test/lisp/emacs-lisp/bytecomp-tests.el (bytecomp-test-var)
(bytecomp-test-get-var, bytecomp-test-identity)
(byte-opt-testsuite-arith-data): Add test cases.
For discussion, see the following thread:
https://lists.gnu.org/r/emacs-devel/2021-01/msg01111.html
* test/lisp/autorevert-tests.el
(auto-revert-test07-auto-revert-several-buffers):
* test/lisp/emacs-lisp/seq-tests.el (test-seq-do-indexed)
(test-seq-random-elt-take-all): Fix errors from using add-to-list on
lexical variables.
* test/lisp/emacs-lisp/cl-lib-tests.el
(cl-lib-defstruct-record): Expect test to succeed when byte-compiled
following change of 2021-01-23 'Fix missing file&line info in
"Unknown defun property" warnings'.
(cl-lib-tests--dummy-function): Remove; no longer needed.
(old-struct): Silence byte-compiler warning about unused lexical
variable.
Before this patch doing:
rm lisp/calendar/calendar.elc
make lisp/calendar/cal-hebrew.elc
would spew out lots of spurious such warnings about a `date` argument,
pointing to code which has no `date` argument in sight. This was
because that code had calls to inlinable functions (taking a `date`
argument) defined in `calendar.el`, and while `date` is a normal
lexical var at the site of those functions' definitions, it was
declared as dynbound at the call site.
* lisp/emacs-lisp/byte-opt.el (byte-compile-inline-expand):
Don't impose our local context onto the inlined function.
* test/lisp/emacs-lisp/bytecomp-tests.el: Add matching test.
(pcase--split-pred, pcase--funcall): Adjust for `not`.
(pcase--get-macroexpander): New function.
(pcase--edebug-match-macro, pcase--make-docstring)
(pcase--macroexpand): Use it.
* lisp/emacs-lisp/radix-tree.el (radix-tree-leaf): Use it!
* doc/lispref/control.texi (The @code{pcase} macro): Document it.
* lisp/emacs-lisp/ert.el (ert--explain-equal-rec): Remove redundant test.
* test/lisp/progmodes/elisp-mode-tests.el (xref-elisp-test-run):
Make sure file names can be compared as strings, by running them
through 'file-truename'. Reported by Vin Shelton
<acs@alumni.princeton.edu>.
* test/lisp/emacs-lisp/bytecomp-tests.el ("warn-obsolete-hook.el")
("warn-obsolete-variable.el"): Use [^z-a] to match a newline as
well. Reported by Vin Shelton <acs@alumni.princeton.edu>.
Implement a limited form of tail-call optimization for the special
case of recursive functions defined with `cl-labels`. Only self-recursion
is optimized, no attempt is made to handle more complex cases such a mutual
recursion.
The main benefit is to reduce the use of the stack, tho in my limited
tests, this can also improve performance (about half of the way to
a hand-written `while` loop).
(cl--self-tco): New function.
(cl-labels): Use it.
* lisp/subr.el (letrec): Optimize single-binding corner case.
* test/lisp/emacs-lisp/cl-macs-tests.el (cl-macs--labels): Add tests
to check that TCO is working.
* lisp/emacs-lisp/eieio-core.el (eieio-defclass-autoload): Try to make
the wording of the warning about the obsoleted variable less confusing.
* lisp/emacs-lisp/bytecomp.el (byte-compile-check-variable): Don't
warn for lexical variables (Bug#39169). Fix spurious `or'.
* test/lisp/emacs-lisp/bytecomp-tests.el
(bytecomp/warn-obsolete-variable-bound\.el): New test.
* test/lisp/emacs-lisp/bytecomp-resources/warn-obsolete-variable-bound.el:
New file.
* test/lisp/emacs-lisp/package-tests.el
(package-test-list-filter-by-name, package-test-list-clear-filter):
Use ansi-color instead of tetris, which no longer has a version:.
* lisp/emacs-lisp/find-func.el (find-function-search-for-symbol): If
our regexp algorithm could not find a location for the symbol
definition, resort to find-function--search-by-expanding-macros.
* test/lisp/emacs-lisp/find-func-tests.el: Add a automatic test for a
function and variable generated by a macro.
* etc/NEWS: Advertise the improved functionality (bug#45443).
* lisp/emacs-lisp/bytecomp.el (byte-compile-file): Don’t fail if
target filename doesn’t contain a directory name.
* test/lisp/emacs-lisp/bytecomp-tests.el
(bytecomp-tests--target-file-no-directory): New unit test.
See Bug#44631. While testing for a readonly output directory has
slightly different semantics, in practice they should cover cases
where Emacs is sandboxed and can only write to the destination file,
not its directory.
* lisp/emacs-lisp/bytecomp.el (byte-compile-file): Handle the case
where the output directory is not writable.
* test/lisp/emacs-lisp/bytecomp-tests.el
(bytecomp-tests--not-writable-directory)
(bytecomp-tests--dest-mountpoint): New unit tests.
The debugger needs to receive a list of the error symbol and data;
cf. the documentation of the `debugger' variable. This bug manifested
itself in ERT forms such as (should (integerp (ert-fail "Boo"))),
which resulted in an incorrect condition object. Note that forms such
as (should (ert-fail "Boo")) weren't affected because they wouldn't
use the `ert--should-signal-hook'.
* test/emacs-lisp/ert.el (ert--should-signal-hook): Call debugger with
the right arguments.
* test/lisp/emacs-lisp/ert-tests.el (ert-test-fail-inside-should): Add
unit test.
