The undo item for ENTER should record the file size *before* the amount
of auto-indentation whitespace is added to it.
This fixes https://savannah.gnu.org/bugs/?54344.
Reported-by: Liu Hao <lh_mouse@126.com>
Bind the until-now unbound function 'cutwordright' to <Ctrl+Delete>.
The complementary function, 'cutwordleft', is not bound by default
because on many terminals the keystroke <Ctrl+Backspace> generates
^H -- the canonical ASCII backspace character. We cannot change the
existing action of ^H without upsetting some users.
Signed-off-by: Marco Diego Aurélio Mesquita <marcodiegomesquita@gmail.com>
Signed-off-by: Benno Schulenberg <bensberg@telfort.nl>
To match the documentation and to match what the nanorc.nanorc
colors as valid.
In theory it would be possible to allow rebinding also F17...F63,
but nano does not recognize escape sequences for these high-order
function keys. As no one ever complained about unknown sequences
when pressing function keys, I am guessing that no one has these
high-order keys, or at least that no one uses them.
This fixes https://savannah.gnu.org/bugs/?54332.
Instead of being entirely silent when ^] is hit after whitespace
or punctuation, report what is lacking -- similar to M-] saying
"Not a bracket" when the cursor is not sitting on a bracket.
This makes the ^] keystroke more discoverable.
This makes things symmetrical: ^W starts a forward search, ^Q starts
a backward search, M-W searches the next occurrence forward, and M-Q
searches the next occurrence backward.
The Tabs-To-Spaces toggle is moved to M-O, and thus the More-Space
toggle is no longer bound by default.
When using option -K on an xterm-like terminal, pressing <Shift+Del>,
<Alt+Del>, or <Ctrl+Del> would enter "2~", "3~", or "5~", respectively,
into the buffer. Now it will just report an "Unknown sequence".
The crawl needs at least five rows to be perceived as a crawl, and
the widest line is 31 characters (ignoring translations, which might
be even longer). Formerly, when only two or three rows were available,
nothing was shown at all, which was puzzling. So, better do the credits
only when there is room enough for them, and print a message otherwise.
When switching to a different buffer, don't just show its name but
also the number of lines it contains. This is useful extra info.
Then use this same message when at startup multiple files are opened
and (after reading them all) we switch back to the first buffer.
(This loses, when multiple files are opened, the information about
format conversion that nano still shows when a single file is opened,
but... this bug has shown that people don't really look at this line
anyway, so... let it be. The info can still be seen when writing out
the file with ^O.)
This addresses https://savannah.gnu.org/bugs/?54047.
Just like an <Esc> before a Ctrl+letter keystroke is ignored, an <Esc>
before an Alt+letter keystroke should be ignored too -- it should not
be interpreted as if the user had typed <Esc> <Esc> letter.
This fixes https://savannah.gnu.org/bugs/?54301.
Even when we don't act on most of these combinations, at least eat up
the whole sequence so the unknown keystroke doesn't enter things like
"3~" into the buffer.
This fixes https://savannah.gnu.org/bugs/?54291.
Unlike glibc, which in getc() locks the file only when it is needed,
FreeBSD and Bionic libc will always lock the file, causing a massive
slowdown, as the system has to create and destroy a mutex each time
getc() is called.
Avoid that massive overhead by locking the file before starting to read
and unlocking it after reading is complete, and using getc_unlocked() to
read each byte. This makes reading on FreeBSD/macOS and Android anywhere
from 2 to 6 times faster, and on glibc roughly seventy percent faster.
This partially addresses https://savannah.gnu.org/bugs/?50406.
Signed-off-by: Devin Hussey <husseydevin@gmail.com>
The 'statusbar_x' variable does not refer to a position in the status
bar, but to a position in the answer that is currently being typed at
the prompt. So... let's call it 'typing_x', for lots of contrast.
And in the bargain get rid of some duplicate code.
This makes a binary without UTF-8 support slightly slower, but that's
not important -- it is more than fast enough anyway. Important is that
the most used and longest code path, the UTF-8 case, becomes faster.
Note that 'is_cntrl_mbchar()' will fall back to 'is_cntrl_char()' for
a non-UTF-8 build, so the deleted piece of code really was equivalent
with the remaining piece for that case.
Take our cue from undoing/redoing line joins: when they take place on
the magicline without --nonewlines, they produce undo/redo items that
don't do anything to the text, but still position the cursor properly.
Reshuffle a bit of code so that we can do the same for magicline cuts.
This fixes https://savannah.gnu.org/bugs/?54032.
Again, if the most significant bit of a UTF-8 byte is zero, it means
the character is a single byte and we can skip the call of mblen(),
*and* if the character is one byte it also occupies just one column,
because all ASCII characters are single-column characters -- apart
from control codes.
This partially addresses https://savannah.gnu.org/bugs/?51491.
For UTF-8, if the most significant bit of a byte is zero, it means the
character is just a single byte and we can skip the call of mblen().
For files consisting of pure ASCII bytes (between 0x00 and 0x7F), this
change reduces the counting time of mbstrlen() by ninety six percent.
This partially addresses https://savannah.gnu.org/bugs/?50406.
When mbtowc() is never called with anything less than MAXCHARLEN as
the length parameter, it will apparently not get confused and will
not need to be reset.
The variable 'last_action' is set at the end of these functions,
serving to prevent a fresh action from being merged with an older
action further down on the stack. Setting 'last_action' before
exiting from undo()/redo() makes no sense.
Also adjust a sideways related old Changelog item.
The macro stuff, however, is basically my code: I told him exactly
how to do it, and I changed most of the patch before applying it.
Also, put myself in second place, as I've just about overtaken David
in the duration of maintainership.
Otherwise the first line of a multiline /*...*/-comment would be
seen as quoted and thus *not* as the first line of a paragraph.
In the code, use "/{2}" to prevent the remainder of the line
getting colored as a comment.
The paragraph-aware indenting is needed only when automatic long-line
wrapping is on (that is, when not using 'nowrap': when writing prose).
When writing code, the user most likely uses --nowrap, which will make
nano indent a new line always to the same amount as the preceding line
-- when --autoindent is in effect, of course.
The only time that feedback about the number of lines written is *not*
wanted is when writing a temporary file.
This fixes https://savannah.gnu.org/bugs/?54025.
When doing autoindentation, and the next line is not the start of
a new paragraph, then use the indentation of that line for the new
line, as it is more likely to have the desired indentation -- the
current line might be the start of the paragraph and thus could
have a deviant indentation.
Pico wraps at the last blank character before or on the target column;
if there is no such blank, then it will wrap at the first blank *after*
the target column -- when it can wrap, it will wrap. Nano should do
the same (and judging from the comments, it intended to do the same),
instead of turning the paragraph into a single unwrapped line.
This fixes https://savannah.gnu.org/bugs/?53986.
First step to the next buffer, and then iterate until we either
find a match or are back at the buffer where we started.
This fixes https://savannah.gnu.org/bugs/?53970.
Signed-off-by: Marco Diego Aurélio Mesquita <address@hidden>
The two help lines will be redrawn in main() only when the current
menu is not MMAIN. So, set it to MLINTER as soon as the help lines
are cleared, just before preparing to invoke the linter.
This fixes https://savannah.gnu.org/bugs/?53967.
Reported-by: Marco Diego Aurélio Mesquita <marcodiegomesquita@gmail.com>
When executing a command in the current buffer and piping this buffer
(or marked region) to that command, then the cutting of the existing
text and the insertion of the new text should be undone and redone
together, as to the user they appear as a single operation.
With-help-from: Marco Diego Aurélio Mesquita <marcodiegomesquita@gmail.com>
When executing a command, it is now possible to pipe the entire buffer
(or the marked region, if anything is marked) to the external command.
The output from the command replaces the buffer (or the marked region),
or goes to a new buffer.
This fulfills https://savannah.gnu.org/bugs/?28993,
and fulfills https://savannah.gnu.org/bugs/?53041.
Signed-off-by: Marco Diego Aurélio Mesquita <marcodiegomesquita@gmail.com>
Signed-off-by: Benno Schulenberg <bensberg@telfort.nl>
In the past, the argument could be either a regex or a literal string,
so the wording was kind of vague. But nowadays we can count on having
regex support (through gnulib), so be more precise in the description.
If the edit window consists of a single row, then the do_up() call has
already brought the desired line into view -- which means that trying
to scroll then (when already on the first line of the file) would fail.
This fixes https://savannah.gnu.org/bugs/?53891.
Letters in other scripts than English often consist of multiple bytes.
When in a UTF-8 locale, correctly gather these bytes from the input and
put them together as the single letter that they are, before comparing
this letter against the translated "Yy", "Nn" and "Aa" strings.
The tiny version contains much less code, so is less likely
to crash. And the users most likely use it for very simple
and short editing jobs, making the chance of a crash still
smaller. So the handler would just be bloat.
When 'afterends' is set and Ctrl+Right or Shift+Ctrl+Right is pressed,
nano will stop at the ends of words instead of their beginnings.
Signed-off-by: Mark-Weston <markweston@cock.li>
Signed-off-by: Benno Schulenberg <bensberg@telfort.nl>
Upon a segmentation fault or an abort signal, instead of crashing,
losing all changes, and leaving the terminal in curses mode, nano
now calls die(), to save any changed buffers and to restore the
terminal to a usable state.
For the remote chance that nano segfaults in die(), the handler for
SIGSEGV and for SIGABRT is reset to its default value as soon as the
signal fires, to prevent a crash-handler loop.
Since a core dump is usually more helpful for debugging, the crash
handler is not included in a debug build.
This addresses https://savannah.gnu.org/patch/?9623.
Signed-off-by: Devin Hussey <husseydevin@gmail.com>
Also, add a period after the "Read nn lines" message, and
don't let the next shell prompt overwrite this message.
This addresses https://savannah.gnu.org/bugs/?53779.
This makes the names of these bindable functions equal to the names of
their corresponding options -- like for all the other toggles that have
a corresponding option.
Set the NO_NEWLINES flag to achieve this. And move the saving and
restoring of the global flags to the main speller routine, so the
flags aren't saved and restored for each internal spell fix.
This fixes https://savannah.gnu.org/bugs/?53742.
Acked-by: David Lawrence Ramsey <pooka109@gmail.com>
When the cursor is on the top or bottom line of the edit window, and
an <Up> or <Down> pushes the cursor offscreen, then use edit_scroll()
to bring it back into view, instead of using edit_redraw(), because
the latter would redraw *every row* on the screen, which is a waste
of time.
This addresses https://savannah.gnu.org/bugs/?53562.
Reported-by: Devin Hussey <husseydevin@gmail.com>
Bug existed since commit 30fc197b (a month ago) which changed the
type of 'i' from int to size_t, causing the comparison to do the
wrong thing when 'threshold' is negative.
This fixes https://savannah.gnu.org/bugs/?53722.
Reported-by: Devin Hussey <husseydevin@gmail.com>
The formatter or linter might have changed in the meantime -- when
the filename was changed to have a different extension, for example.
This fixes https://savannah.gnu.org/bugs/?53716.
When the user configures nano with --enable-altrcname=name,
the specified name should not be prefixed with a dot.
This fixes https://savannah.gnu.org/bugs/?53694.
Reported-by: Cody A. Taylor <codemister99@yahoo.com>
When the keyboard buffer is empty, return a value that is not equivalent
to <Ctrl+Space>, so that typing a bunch of these while searching is going
on will not result in some of them getting executed afterward.
Indenting/unindenting always happens in a single contiguous block.
Lines that cannot or should not be indented or unindented do not
need to remembered separately as it follows from their content.
That is, do not fall back to the default background color when the
name for the background color is invalid, but reject the entire
color command, just like for an invalid foreground color.
In the preceding commit, open_buffer() was changed so that it gets
told whether to load into a new buffer or not, so it is no longer
needed to convey this information through a flag.
This fixes a regression introduced by commit 54103d8e: a crash that
can be triggered by running 'nano --restrict' and pressing <Insert>.
This addresses https://bugzilla.redhat.com/1558532.
When the file has just been read, it is in the state that it has
on disk, so there is nothing to undo.
This fixes https://savannah.gnu.org/bugs/?53387.
When the cursor is on the first or last row of the edit window, and
thus Scroll-Down or Scroll-Up would push it offscreen, first move
the cursor away from the edge row and then scroll.
This fixes https://savannah.gnu.org/bugs/?53376.
As, since commit 0e30177d, the scrollup and scrolldown commands
no longer intend to move the cursor, they should not be seen as
movement functions.
Also, it is not guaranteed that functions are ordered in the same
way in the binary as in the source code, so an ordered comparison
of function pointers will not always work.
This fixes https://savannah.gnu.org/bugs/?53333.
When just scrolling and the cursor does not need to change position
(that is: it is not on the first or last row of the edit window),
then edit_scroll() has handled everything and there is no need to
additionally redraw anything or update 'placewewant'.
Instead of keeping the cursor in the same spot on the screen,
let the cursor move with the text (whenever possible).
This makes these functions behave the same as in Vim and Emacs.
This fixes https://savannah.gnu.org/bugs/?53269.
Reported-by: Ashish Kunwar <dorkerdevil280@gmail.com>
The leak was mistakenly introduced by commit dc3618a1
-- probably as a leftover of testing things.
Help texts are shown without any line numbers, so the full width
of the screen should be used to display the text.
This fixes https://savannah.gnu.org/bugs/?53308.
First the two that add something (ADD, ENTER), then the three that
delete something (BACK, DEL, JOIN), and then the one that changes
something (REPLACE). Then the SPLITs, CUT, PASTE, and INSERT, and
then the INDENTs and COMMENTs, when they exist.
For cuts, pastes, and inserts, the lines have already been renumbered;
for indents, comments, and replacements, the line numbers cannot have
changed. (And anyway, variable 'f' is not set for those cases.)
Only when lines get split (ENTER) or fused together (JOIN) do the later
lines need to be renumbered. This mirrors what is done for do_redo().
The 'f' variable is used only in the ADD, BACK, DEL, ENTER, JOIN, and
REPLACE undo/redo cases. So, avoid making a somewhat costly call when
it is entirely superfluous. Rearrange the undo types to make checking
for the above six types easier.
This assumes that all movement functions are located together,
starting with to_first_line() and ending with do_right().
This fixes https://savannah.gnu.org/bugs/?53195.
Reported-by: Brand Huntsman <alpha@qzx.com>
The implant() function itself cannot be bound to anything, so it
is not in the linked list of functions -- trying to find it would
lead to a NULL pointer, and thus to a segfault.
This fixes https://savannah.gnu.org/bugs/?53233.
Instead of always calling sctofunc(), it is now only called when
in view mode OR when (after the keystroke's function has been run)
no need for a refresh has been established yet.
Commit bb667beb removed do_gotolinecolumn_void() from the MWHEREIS
menu (replacing it with the empty flip_goto() function), so there
is no longer any need to check for this one special case.
The slash is easier to read than the underscore (which almost
disappears at the bottom of the screen), and easier to type
(no Shift needed on a US keyboard), and it kind of harmonizes
with the ^\ for Replace and the M-/ for End-of-buffer.
When using --smooth or 'set smooth', the screen should scroll the
minimum amount needed to get the cursor back into view. (The only
exceptions are search, undo, and redo -- when there the cursor goes
offscreen, the cursor line is centered.)
This change brings the behavior of pressing <Enter> on the bottom
row into line with, for example, pasting a single line. See also
http://lists.gnu.org/archive/html/nano-devel/2018-02/msg00027.html.
When 'refresh_needed' is already TRUE, there is no need any more
to check whether it should be set.
[Those first two calls are leftovers from before the time that
reset_multis() morphed into check_the_multis().]
When characters are added via Verbatim Input or by pressing <Tab>,
the prepend flag should be retained, just like when characters are
typed directly.
This fixes https://savannah.gnu.org/bugs/?52956.
When copying a string, source and destination may not be equal --
complain loudly when they are, instead of failing to free memory.
Also, instead of freeing the destination string and then allocating
it afresh, just reallocate it -- that should be slightly quicker.
When doing for example: ^W xx ^R ^C ^W, the "xx" would again be shown
after the prompt. This is wrong -- when starting a new search, the
current answer should be empty.
Reported-by: Marco Diego Aurélio Mesquita <marcodiegomesquita@gmail.com>
Instead of weaving back and forth between do_search()/do_replace() and
search_init() every time that one of the options is toggled, just keep
looping in the latter function until the user presses <Enter>.
The weaving for the do_gotolinecolumn() function remains, but is
shortened to involve only search_init().
Benno Schulenberg