This lets Emacs deal with arbitrary local symlinks without
mishandling their contents (Bug#28156). For example,
(progn (shell-command "ln -fs '~' 'x'") (rename-file "x" "/tmp/x"))
now consistently creates a symbolic link from '/tmp/x' to '~'.
Formerly, it did that only if the working directory was on the
same filesystem as /tmp; otherwise, it expanded the '~' to
the user's home directory.
* lisp/dired.el (dired-get-filename): Use files--name-absolute-system-p
instead of rolling our own code.
* lisp/files.el (files--name-absolute-system-p): New function.
(file-truename, file-chase-links): Use it to avoid mishandling
symlink contents that begin with ~.
(copy-directory, move-file-to-trash):
Use concat rather than expand-file-name, to avoid mishandling
symlink contents that begin with ~.
* src/fileio.c (Fmake_symbolic_link): Do not expand leading "~" in the
target unless interactive. Strip leading "/:" if interactive.
(emacs_readlinkat): Do not prepend "/:" to the link target if
it starts with "/" and contains ":" before NUL.
* test/src/fileio-tests.el (try-link): Rename from try-char,
and accept a string instead of a char. All uses changed.
(fileio-tests--symlink-failure): Also test leading ~, and "/:",
to test the new behavior.
* doc/lispref/files.texi (File Attributes):
Document file-attributes atomicity.
* etc/NEWS: Document the fix.
* src/dired.c (file_attributes): New args DIRNAME and FILENAME,
for diagnostics. All callers changed. On platforms like
GNU/Linux that support O_PATH, fix a race condition in
file-attributes and similar functions, so that these functions do
not return nonsense if a directory entry is replaced while getting
its attributes. On non-GNU platforms, do a better (though not
perfect) job of detecting the race, and return nil if detected.
* src/fileio.c (expand_and_dir_to_file): Simplify by omitting 2nd
argument, since in practice it always has the default value. All
callers changed. Prefer C99 style decls in nearby code.
* src/xterm.c (xm_scroll_callback, xaw_jump_callback)
(x_set_toolkit_scroll_bar_thumb)
(x_set_toolkit_horizontal_scroll_bar_thumb): Prefer ‘double’ to
‘float’ for individual local floating-point temporaries.
* src/nsimage.m (ns_load_image): Remove index check.
(EmacsImage::getAnimatedBitmapImageRep): New function.
(EmacsImage::getMetadata): Use getAnimatedBitmapImageRep.
(EmacsImage::setFrame): Use getAnimatedBitmapImageRep and check index
is valid.
* src/intervals.c (get_local_map): Don't allow C-g to quit as long
as we have the buffer widened, to make sure the restriction is
preserved. (Bug#28161)
* lisp/net/tramp.el (tramp-interrupt-process): Rename from
`tramp-advice-interrupt-process'. Adapt according to changed API.
(top): Add it to `interrupt-process-functions'.
* src/process.c (Finternal_default_interrupt_process): New defun.
(Finterrupt_process): Change implementation, based on
Vinterrupt_process_functions.
(Vinterrupt_process_functions): New defvar.
* test/lisp/net/tramp-tests.el (tramp-test40-unload): Do not
test removal of advice.
* lisp/files.el (make-temp-file): Let make-temp-file-internal do
the work of inserting the text.
* src/fileio.c (Fmake_temp_file_internal): New arg TEXT.
All callers changed.
* src/w32term.c (x_get_scale_factor): New function.
(w32_draw_underwave): Use it.
* src/xterm.c (x_draw_underwave): Offset the wave starting point
to make it identical with original code.
* src/xterm.c (x_draw_underwave): Compute height, length and thickness
based on scale factor.
(x_get_scale_factor): New function.
Copyright-paperwork-exempt: yes
This incorporates:
2017-08-15 renameat: ensure declaration in <stdio.h> on NetBSD
2017-08-15 extensions: enable NetBSD specific extensions
2017-08-14 open: support O_CLOEXEC
2017-08-13 reallocarray: new module
* admin/merge-gnulib (AVOIDED_MODULES): Remove ‘open’, since
it now supports O_CLOEXEC and this simplifies Emacs.
* build-aux/config.guess, lib/fcntl.in.h, lib/stdio.in.h:
* lib/stdlib.in.h, m4/extensions.m4, m4/stdlib_h.m4:
Copy from Gnulib.
* lib/cloexec.c, lib/cloexec.h, lib/open.c:
* m4/mode_t.m4, m4/open-cloexec.m4, m4/open.m4:
New files, copied from Gnulib.
* lib/gnulib.mk.in, m4/gnulib-comp.m4: Regenerate.
* lib-src/etags.c (O_CLOEXEC) [WINDOWSNT]:
Remove, as Gnulib does this for us.
* src/filelock.c (create_lock_file):
* src/sysdep.c (emacs_open, emacs_pipe):
Don’t worry about O_CLOEXEC == 0, as Gnulib no longer sets it to 0.
* src/nsterm.h (NSWindowTabbingMode): Define in pre-Sierra macOS.
(MAC_OS_X_VERSION_10_6, MAC_OS_X_VERSION_10_7, MAC_OS_X_VERSION_10_8,
MAC_OS_X_VERSION_10_9, MAC_OS_X_VERSION_10_12, HAVE_NATIVE_FS): Remove
defines.
(NSWindowStyleMaskFullScreen,
NSWindowCollectionBehaviorFullScreenPrimary,
NSApplicationPresentationFullScreen,
NSApplicationPresentationAutoHideToolbar): Define in macOS 10.6.
* src/nsterm.m (colorForEmacsRed, colorUsingDefaultColorSpace,
check_native_fs, ns_read_socket, ns_select, runAlertPanel,
initFrameFromEmacs, windowDidMiniaturize, windowDidEnterFullScreen,
windowDidExitFullScreen, isFullscreen, updateCollectionBehavior,
toggleFullScreen, constrainFrameRect, scrollerWidth, syms_of_nsterm):
Allow use of run-time checks and replace version check macros.
* src/nsfns.m (ns_screen_name): Use run-time OS version checks.
* src/macfont.m (macfont_draw): Use run-time OS version checks.
* src/nsmenu.m (menuWillOpen): Use run-time OS version checks.
Co-authored-by: Charles A. Roelli <charles@aurox.ch>
* src/fileio.c (Frename_file): On macOS, renameat_noreplace can
fail with errno == ENOTSUP on file systems where it is not
supported, according to the Apple documentation.
Problem reported by Philipp Stephani (Bug#27986).
* src/fileio.c (Frename_file):
Worry about file name case sensitivity only if CYGWIN or DOS_NT.
* src/sysdep.c (renameat_noreplace): Use renameatx_np on macOS,
since this provides the necessary atomicity guarantees.
* src/indent.c (Fvertical_motion): If lines are truncated and we
end up beyond the right margin of the window, don't assume we are
in the next screen line, unless VPOS actually says so. (Bug#28071)
The bug with "." and ".." has been present for a while; I
introduced the bug with "" earlier today in my patch for Bug#28023.
* lisp/files.el (make-temp-file): Do not use expand-file-name if
PREFIX is empty or "." or "..", as it does the wrong thing.
Compute absolute-prefix here ...
(files--make-magic-temp-file): ... instead of here ...
* src/fileio.c (Fmake_temp_file_internal): ... or here.
* lisp/files.el (make-temp-file): If the prefix is empty, append
"/" to the absolute prefix so that the new files are children
rather than siblings of temporary-file-directory. This fixes a
bug introduced in the previous change.
* test/lisp/files-tests.el (files-test-make-temp-file-empty-prefix):
New test, for the bug.
For the motivation behind this patch, please see Bug#28023 and:
http://emacshorrors.com/posts/make-temp-name.html
Although, given the recent changes to Tramp, the related security
problem in make-temp-file is already fixed, make-temp-file still has
several unnecessary system calls. In the typical case on GNU/Linux,
this patch replaces 8 syscalls (symlink, open, close, readlinkat, uname,
getpid, unlink, umask) by 2 (open, close).
* admin/merge-gnulib (GNULIB_MODULES): Add tempname, now
that Emacs is using it directly.
* configure.ac (AUTO_DEPEND): Remove AC_SYS_LONG_FILE_NAMES;
no longer needed.
* lib/gnulib.mk.in, m4/gnulib-comp.m4: Regenerate.
* lisp/files.el (files--make-magic-temp-file): Rename from
make-temp-file.
(make-temp-file): Use make-temp-file-internal for
non-magic file names.
* src/fileio.c: Include tempname.h.
(make_temp_name_tbl, make_temp_name_count)
(make_temp_name_count_initialized_p, make_temp_name): Remove.
(Fmake_temp_file_internal): New function.
(Fmake_temp_name): Use it.
* src/filelock.c (get_boot_time): Use Fmake_temp_file_internal
instead of make_temp_name.
* src/msdos.c (faccessat):
* src/w32.c (faccessat): Support relative file names, and add D_OK
to 'mode' if the argument is a directory. This unbreaks file-name
completion when the completion result is a directory.
* lib-src/ntlib.h (mkdir, open): Remove redefinitions. They are
now in nt/inc/ms-w32.h.
* lib-src/ntlib.c (sys_mkdir, sys_open): New functions.
(mkostemp): Remove.
* src/w32.c (mkostemp): Remove.
(sys_mkdir): Accept a second (unused) argument.
* src/fileio.c (Fmake_directory_internal): Remove the WINDOWSNT
specific call to mkdir. (Bug#28023)
* nt/inc/ms-w32.h (mkdir): Remove from "#ifdef emacs" and redefine
to accept 2 arguments.
(open): Remove from "#ifdef emacs".
* nt/mingw-cfg.site (ac_cv_func_mkostemp): Remove.
* nt/gnulib-cfg.mk (OMIT_GNULIB_MODULE_mkostemp)
(OMIT_GNULIB_MODULE_tempname): Remove.
Although this does not fix Bug#27986, it is a step forward.
I plan to propose a more-significant patch later.
* lisp/files.el (directory-name-p): Move from here ...
* src/fileio.c (Fdirectory_name_p): ... to here.
(directory_like, cp_like_target): New static functions.
(Fcopy_file, Frename_file, Fadd_name_to_file)
(Fmake_symbolic_link):
Use them, to avoid directory-testing syscalls on file names that
must be directories if they exist. Omit unnecessary
initializations and CHECK_STRING calls.
(Frename_file): Don't call file_name_case_insensitive_p
twice on the same file. Compare both file names expanded, instead
of the old name expanded and the new one unexpanded.
* src/w32fns.c (syms_of_w32fns) <w32-disable-abort-dialog>: New
variable.
(emacs_abort): If w32-disable-abort-dialog is non-nil, abort right
away, without displaying the Abort dialog, which waits for the user.
* test/src/emacs-module-tests.el (module--test-assertion): Run the
inferior Emacs with the w32 abort dialog disabled. Expect the
status of the aborted Emacs sub-process to be 3 on MS-Windows and
2 on MS-DOS.
* doc/lispref/files.texi (Unique File Names):
* src/fileio.c (Fmake_temp_name): Document that make-temp-name
does not guarantee uniqueness on magic file names.
* src/fns.c (syms_of_fns): New variable `overriding-plist-environment'.
(Fget): Consult it.
* lisp/emacs-lisp/bytecomp.el (byte-compile-close-variables): Let-bind
it to nil.
(byte-compile-define-symbol-prop): New function, handles compilation
of top-level `define-symbol-prop' and `function-put' calls by putting
the symbol setting into `overriding-plist-environment'.
Co-authored-by: Noam Postavsky <npostavs@gmail.com>
* src/xdisp.c (get_overlay_strings_1)
(handle_single_display_spec, push_prefix_prop): Invalidate the
composition information before starting to iterate on a string.
Otherwise we might think in set_iterator_to_next that we are
delivering characters from a composition, and do all kinds of
nonsensical things, like over-step the string end. (Bug#27761)
Prompted by a problem report by Alex in:
http://lists.gnu.org/archive/html/emacs-devel/2017-08/msg00143.html
* src/eval.c (For, Fprogn, Fsetq, FletX, eval_sub):
Compute XCDR (x) near XCAR (x); although this doesn't fix any bugs,
it is likely to run a bit faster with typical hardware caches.
(Fif): Use Fcdr instead of XCDR, to avoid crashing on
self-modifying S-expressions.
(Fsetq, Flet, eval_sub): Count the number of arguments as we go
instead of trusting an Flength prepass, to avoid problems when the
code is self-modifying.
(Fquote, Ffunction, Fdefvar, Fdefconst): Prefer !NILP to CONSP
where either will do. This is mostly to document the fact that
the value must be a proper list. It's also a tiny bit faster on
typical machines nowadays.
(Fdefconst, FletX): Prefer XCAR+XCDR to Fcar+Fcdr when either will do.
(eval_sub): Check that the args are a list as opposed to some
other object that has a length. This prevents e.g. (if . "string")
from making Emacs dump core in some cases.
* test/src/eval-tests.el (eval-tests--if-dot-string)
(eval-tests--let-with-circular-defs, eval-tests--mutating-cond):
New tests.
Respect comment-auto-fill-only-comments when auto-filling and a
comment syntax is defined.
* lisp/newcomment.el (comment-indent-new-line): Do not check
comment-auto-fill-only-comments.
* lisp/simple.el (internal-auto-fill): New defun.
* src/cmds.c (internal_self_insert): Call Qinternal_auto_fill, not
auto_fill_function.
(syms_of_cmds): Define Qinternal_auto_fill.
* src/buffer.c (Fgenerate_new_buffer_name): Test IGNORE for being
nil before calling string-equal, since the latter will compare
"nil and 'nil' as equal. (Bug#27966)
* test/src/buffer-tests.el
(test-generate-new-buffer-name-bug27966): New test.
Problem reported by Glenn Morris (Bug#27708#58).
* src/gnutls.c (HAVE_GNUTLS_X509_SYSTEM_TRUST):
New macro. Use it instead of low-level version number checks.
(HAVE_GNUTLS_AEAD): Move here from gnutls.h, and rename from
HAVE_GNUTLS3_AEAD. All uses changed. Indent preprocessor lines.
* src/gnutls.h (HAVE_GNUTLS3_CIPHER, HAVE_GNUTLS3_DIGEST)
(HAVE_GNUTLS3_HMAC): Remove, since these were available
before GnuTLS 3.0.0 and the code checks them only if HAVE_GNUTLS3
is defined. Remove all uses; this simplifies the code a bit.
Problem reported by Ted Zlatanov in:
http://lists.gnu.org/archive/html/emacs-devel/2017-08/msg00082.html
* src/fileio.c (Frename_file): On RHEL 7 + NFS, renameat2 can fail
with errno == EINVAL when it is not supported. So treat that case
like errno == ENOSYS. Also, when ok_if_already_exists is neither
nil nor an integer, just call plain rename; this avoids an extra
syscall to renameat2 when the latter fails with errno == EINVAL or
ENOSYS or ENOENT.
There's only one GnuTLS, so configuring these symbols at
'configure' time is overkill. Simplify things by moving their
configuration to src/gnutls.h (Bug#27708).
* configure.ac (HAVE_GNUTLS3, HAVE_GNUTLS3_HMAC, HAVE_GNUTLS3_AEAD)
(HAVE_GNUTLS3_CIPHER, HAVE_GNUTLS3_DIGEST): Move these definitions
from here ...
* src/gnutls.h: ... to here, and simplify.
