Do the same what the normal "research" code does: if nothing was
searched for yet during this session, and there is a history of
search items, then search for the last item in that history.
Make case-insensitive searching in a UTF-8 locale eight times faster
when the actual characters involved are plain ASCII.
This makes us faster than 'less', and as fast as Vim and Emacs.
The disadvantage of this change is that searching for a string that
begins with a multibyte character is nearly ten times slower than
searching for one that begins with an ASCII character. This may be
unsettling when searching a huge file first for a simple ASCII string
and later for a UTF-8 one. Doing this second search, the user might
get impatient: "Why is it taking so long?"
(This patch fell through the cracks four years ago, when I worked on
the searching code. It sat in a branch on top of other changes that
I never applied because I made different improvements. The speedup
at the time, on that machine, was only around sixty percent, though.
But measuring it now again on the same machine, it clocks in at an
82 percent reduction with -O0 and an 87 percent reduction with -O2.)
Those casts are redundant, and sometimes ugly. And as the types of
variables are extremely unlikely to change any more at this point,
the protection they offer against miscompilations is moot.
Signed-off-by: Hussam al-Homsi <sawuare@gmail.com>
The two functions findnextstr() and do_replace_loop() do not change
or even touch 'last_search', so there is no need to save and then
restore its value when doing corrections of misspelled words.
Storing the orientation of the marked region beforehand is not needed,
as this orientation is readily available also after the justification.
(By the way, cursor and mark need to be swapped after justifying
a backward-marked region because the rule is that the cursor gets
placed *after* the justified paragraph. Maybe that should change?)
The most likely reason for stat() returning -1 is that the file
does not exist. And an absent positionlog file is not an error.
(In some cases it is, like immediately after writing the file,
but even then we don't want to complain, because it may have
been some other process that deleted the file straightaway.)
This fixes https://savannah.gnu.org/bugs/?58993.
Bug existed since version 5.0, commit fcb9e58b.
This ability was lost in commit 92298349 from two hours ago, which
bypasses the keystroke buffer and its integrated screen resizing.
This new implementation is better than it was before, because it
responds almost instantly to a resize instead of with a delay of
up to a second.
(Still, this does not allow a full escape sequence to be used as
the Cancel command, but I think that is an acceptable limitation,
because 1) nobody ought to be using --rawsequences, and 2) very
few people will bind Cancel to something like F3 or Ins.)
This improves the fix for https://savannah.gnu.org/bugs/?58825.
When checking (during a Search command) whether the user has pressed
the Cancel keystroke, look at ncurses' input stream directly instead
of at nano's own keystroke buffer, because the latter may contain the
copied keystrokes of a macro and we don't want to discard those.
(This does not yet allow a Meta keystroke to be used for Cancel, but
the next commit will fix that.)
This fixes https://savannah.gnu.org/bugs/?58825.
Bug existed since version 2.9.0, since the macro was introduced.
There is no reason to deselect the region, as nothing has changed.
(This also retains the shift-selected region when a non-shortcut key is
typed in view mode, which makes sense, as again nothing was changed.)
This fixes https://savannah.gnu.org/bugs/?58980.
Bug existed since version 4.9, commit 0ed62e84.
When keypad() is set to FALSE, like for verbatim input, ncurses is
not waiting its fifty milliseconds after an ESC to see if another
code will follow it, so nano itself will have to pause a little.
Otherwise 'solitary' could get set to TRUE when in fact the ESC
is followed very closely by another code.
This fixes https://savannah.gnu.org/bugs/?58955.
Bug existed since version 2.6.2, commit f2150d3f.
Also when each character is just one byte, parse_verbatim_kbinput()
can produce two bytes; plus the terminating NUL that is three bytes.
This fixes https://savannah.gnu.org/bugs/?58959.
Bug existed since version 4.9, commit 03d296eb.
This will not work for the deviant escape sequences for F1 to F5
on the Linux console nor for Alt+arrow on urxvt and such, but...
I can't be bothered to handle those too.
This fixes https://savannah.gnu.org/bugs/?58929.
Bug existed since commit be203832 from earlier today.
During verbatim input at most four integers are produced (the longest
possible unicode sequence), so use the value 999 to indicate a special
condition (a screen resize) that should not enter anything into the
buffer AND should not produce any error message or beep.
This fixes https://savannah.gnu.org/bugs/?58923.
Bug existed since version 5.0, commit 5899181a.
When Unicode Input has started (by typing 0 or 1 at the Verbatim Input
"prompt"), and something is typed that is not a hexadecimal digit, then
don't try to enter this character into the buffer but simply report it
as invalid and ignore it. Because most likely the user mistyped and
actually meant to enter a valid hex digit.
This fixes https://savannah.gnu.org/bugs/?58927.
The bug was old -- it existed since at least version 2.0.6.
Just like <Ctrl+Down> goes to the end of the buffer when after the
cursor there are only blank lines.
This fixes https://savannah.gnu.org/bugs/?58921.
Bug existed since version 2.6.2, since <Ctrl+Up/Down> were introduced.
The correct sequences begin with "Esc [" instead. I cannot find in the
histories of VTE and Gnome Terminal when these sequences were corrected.
Also exclude a bug check from the tiny version.
When we get an ESC from the keyboard, it might be the start of an
escape sequence, but the keyboard routines will need a little time
(tens of microseconds, probably) to get these codes to ncurses.
So, when doing verbatim input, pause a moment after an ESC.
This completes the fix for https://savannah.gnu.org/bugs/?58909.
Always interpret ESC followed by BS or DEL as <Alt+Backspace>.
This fixes https://savannah.gnu.org/bugs/?58910.
Bug existed since commit 40b03162 from one week ago,
since M-Bsp was hard-bound to deleting a word backwards.
(This could lead to unwanted behavior when the user wants to enter
a literal escape character (0x1B) while the input is somehow getting
stalled, because then a command keystroke after the M-V <Esc> would
get inserted verbatim too, instead of getting acted upon. But that
is a small price to pay for getting the correct behavior for both
M-V M-Bsp and M-V M-Enter.)
This fixes https://savannah.gnu.org/bugs/?58909.
Bug existed since version 2.6.3, commit 08c51cfd,
but also occurred erratically beforehand.
After commit 59bbc0b8 from five days ago (that made a divergent pipe
failure to be treated in the same way as the other two), this became
possible without having two different pipe-failure messages.
The user had better avoid using the keystroke in a macro, but
when typing directly at a responsive machine there is no problem,
M-[ is a perfectly valid and recognizable keystroke.
This fulfills https://savannah.gnu.org/bugs/?58820.
The make_new_node() function initializes 'extrarows' to a bad value
on purpose, to catch cases like this one where it doesn't get updated
properly.
This fixes https://savannah.gnu.org/bugs/?58890.
Reported-by: Jerry Kindall <kindall@gmail.com>
Bug existed since version 5.0, since the indicator was introduced.
More generally, after invoking the formatter, the entire buffer should
be cut (to be replaced with the reformatted text) and not just the part
before the mark, because then everything after it would be duplicated.
This fixes https://savannah.gnu.org/bugs/?58879.
Bug existed since version 4.9, commit 19517fc5.
When compiling with GCC 10.1.0 and -O2 -Wall, the strncpy() call
in measured_copy() produced two stringop-truncation warnings.
This addresses https://savannah.gnu.org/bugs/?58861.
Signed-off-by: Michalis Kokologiannakis <michalis@mpi-sws.org>
Between an endwin() and a doupdate() there should be no calls of
statusline() or statusbar() -- these two functions may be called
only when in curses mode.
This fixes https://savannah.gnu.org/bugs/?58868.
Bug existed since version 4.3, commit 57390cff.
If forking fails, we must first return to curses mode before we can show
an error message on the status bar. (This additionally requires storing
the error number, because doupdate() apparently sets it to zero.)
This fixes https://savannah.gnu.org/bugs/?58864.
Bug existed since version 4.8, commit 61197563.
It jars a bit that it is <Ctrl+Delete> that deletes a word rightward
and <Alt+Backspace> that deletes a word leftward. But it's good to
also have a two-key keystroke bound by default to 'chopwordleft',
and not just the three-key <Ctrl+Shift+Delete>.
This fulfills https://savannah.gnu.org/bugs/?58709.
Requested-by: Axel Scheepers <axel.scheepers76+gnu@gmail.com>
While ignoring permission errors from fchmod() and fchown() is okay
(since normal users are not always privileged to make such changes),
ignoring also more serious errors (like EIO) is not ideal.
Signed-off-by: Michalis Kokologiannakis <michalis@mpi-sws.org>
When there is just one match when <Tab> is pressed, and this match
is a directory, then a slash is added to 'shared', overwriting the
terminating NUL character. So, strcpy() cannot be used for copying
this 'shared' string, but strncpy() is needed, and the result needs
to be NUL terminated afterward.
This fixes https://savannah.gnu.org/bugs/?58826.
Bug existed since commit b0f56398 from eleven days ago.
The whiles are unneeded because the result of get_input() can
never be NULL when in waiting mode -- and only when searching
for something are we NOT in waiting mode.
<Ctrl+numeric slash> does not invoke Go To Line, so it is fitting
that <Esc> <Esc> <numeric slash> does neither. If some user does
want the double escape plus slash to work, they can use --raw.
(Also, if someone really uses <Esc> <Esc> <numeric slash>, they
probably also type the desired line number on the numeric keypad,
and for that to work they needed to have NumLock engaged, and in
that case the double escape plus slash will work fine.)
(Getting rid of this numeric-slash support makes the three-escapes
case very similar to the one-escape case, allowing the first to be
folded into the latter in the next commit.)
With the two situations that need to preserve the escape counter
now returning directly, the resetting of this counter can happen
at the end of each case block.
The escape-parsing routine nowadays returns FOREIGN_SEQUENCE instead
of ERR when encountering an unknown sequence. So, with the two cases
that demand ERR now handled directly, there is no need to check for
ERR any more.
There is no need to reset the digit counter (because this counter
matters only when the escape counter equals two) -- resetting the
escape counter is enough.
The digit counter will be reset to zero the next time when <Esc>
is pressed, and the byte holder gets re-initialized when the next
first digit is pressed. And the escape counter is reset to zero
after returning from a result-giving call of get_byte_kbinput().
When get_byte_kbinput() is called, it has already been determined that
the keyboard input is a digit from zero to nine (or from zero to two
for the first digit), so there is no need to check this again.
When <Esc><Esc> is followed by digit 0 or 1 or 2 but then NOT two more
digits follow, then both the escape counter and the digit counter should
start afresh when a new ESC code arrives.
This fixes https://savannah.gnu.org/bugs/?58788.
Bug existed since before version 2.0.6.
Root has carte blanche to both read unreadable files and write
unwritable files -- the file system does not prevent this. So,
alert root when opening a file that is intended to be read-only.
This fulfills https://savannah.gnu.org/bugs/?58685.
Requested-by: Marius Bakke <mbakke@fastmail.com>
Also, remove an unneeded 'if', as parse_escape_sequence() is only
ever called when there are at least two bytes after the Esc code.
(If there were not, the 'for' loop after calling convert_sequence()
would use an uninitialized 'consumed' value.)
Rereading is unlikely to fail, but *if* it fails, maybe there is a
serious problem and the user wants to try and fix it before saving
the buffer and thus overwriting the original file.
After thinking about it, I don't like it that nano makes an unrequested
failsafe backup: in the vast, vast majority of cases it is a waste both
of time and of disk wear. If the user is worried about data loss, they
can use --backup. The fsync() after writing out the buffer (added in
commit a84cdaaa) already reduces the chances on data loss somewhat.
This change moves the code responsible for backup creation to a new,
separate function, backup_file(). This function returns a boolean
indicating whether saving the buffer should proceed or not.
Signed-off-by: Michalis Kokologiannakis <michalis@mpi-sws.org>
The escape-sequence codes for the different keypad keys are simply
the relevant ASCII code plus 0x40, so they are easy to translate.
Also, do not accept keypad keys with multiple modifiers (Shift+Alt,
Shift+Ctrl, Alt+Ctrl, Shift+Alt+Ctrl) -- report "Unknown sequence"
for those.
Until now, on an xterm, Shift plus any keypad key would say
"Unknown sequence", which is not useful. On other terminal
emulators (urxvt, Pantheon, LXTerminal), Shift plus a keypad
key already produces the relevant digit or symbol, so this
change harmonizes the behavior of the different emulators.
However, on a Linux console, Shift plus a keypad key moves
the cursor and selects the text that is passed over. And
holding Alt while pressing a keypad key does nothing at all.
For the operator keys on the numeric keypad (/, *, -, and +),
make both the Alt and the Ctrl modifier a no-op.
This fixes https://savannah.gnu.org/bugs/?58731.
Bug existed since version 2.9.3, commit e734488c.
The file-saving procedure that nano followed was not crash-safe
under ext4 (and perhaps other filesystems) when editing existing
files. Specifically, modifying an existing file could lead to data
loss, since a modified file is not replaced atomically on disk.
Using "-B" did not ensure crash-safety either since the backup might
not have persisted on disk when the writeout of the file started.
In addition, if I/O pressure filled up the remaining disk space
after an existing file is truncated during save, nano would not
be able to finish saving the file, which would remain truncated.
This change addresses these issues by making nano do the following:
1) Make a temporary backup of the file being written so that there
is no time window such that (a) an existing file is truncated, and
(b) the buffer corresponding to said file has not been flushed to
disk. Such time windows are dangerous because, under certain
circumstances, they can lead to data loss on some filesystems.
The backup is made by copying the original file, and a second
attempt to HOME is made in case that first copy fails.
2) Use fsync() so that, when the save finishes, it is certain
that the file has been successfully written to disk.
This addresses https://savannah.gnu.org/bugs/?58642.
Signed-off-by: Michalis Kokologiannakis <michalis@mpi-sws.org>
This allows pasting six hexadecimal digits after typing M-V
in order to enter a specific Unicode character.
This fixes https://savannah.gnu.org/bugs/?58730.
Bug existed since version 4.8, since bracketed pastes
were introduced in commit f705a967.
The codes from (for example) U+D800 to U+DFFF are invalid. If any
such invalid code is entered at the Unicode Input prompt, nano should
not act as if the code was accepted and inserted into the buffer, and
should certainly not mark the buffer as Modified.
This fixes https://savannah.gnu.org/bugs/?58714.
Bug existed since version 4.9, commit b3faf353.
There is no need to place the cursor up front, as the drawing of
each name includes the coordinates where the name must be drawn.
Also, trim two comments and reshuffle a free().
When using --constantshow and --nohelp and opening a new buffer,
the cursor location should be shown on the status bar right away.
Bug existed since commit 2e688640 from a few hours ago.
The constant cursor display must be suppressed whenever a message
was printed to the status bar. That is: whenever 'lastmessage' is
something other than 'VACUUM'.
By default, this function is bound to ^L, to make that keystroke do
something actually useful. To not lose the Refresh function that this
keystroke had, the centering function additionally does a full redraw
and refresh of the screen.
Before version 2.9.8, nano would consider any line that started with
leading whitespace as the beginning of a new paragraph, and would thus
be unable to justify an indented piece of text (unless autoindent was
switched on). I thought that this was too limiting and changed the
way nano detects and parses paragraphs. It works fine in texts where
paragraphs occupy multiple lines, but breaks down for paragraphs of
a single line: nano merges those with adjoining lines.
Now, when -O or --bookstyle or 'set bookstyle' is used, nano will again
consider a line that begins with whitespace as the start of a paragraph
-- when 'autoindent' is OFF.
This addresses https://savannah.gnu.org/bugs/?57402.
Indirectly-reported-by: Sébastien Desreux <seb@h-k.fr>
Also-reported-by: Bill Kendrick <nbs@sonic.net>
Instead of calling a function that tries to find the relevant operation,
just add three hidden hard-bindings at the end of the shortcuts list.
They are hidden because we don't want them to show up in the help text,
because we don't know what to call them (as a Cancel and Suspend key
don't exist on normal keyboards) or naming the key would be confusing
(Sh-Ins is grabbed by the desktop manager and doesn't reach nano).
The get_history_completion() function is called in just one place,
and gets called only when the length parameter is larger than zero.
Also, improve a few comments.
Especially when it are just a handful of possible completions, it is
much nicer to have them listed close to the prompt bar where the eyes
of the user are.
This fulfills https://savannah.gnu.org/bugs/?58620.
This gives quicker feedback, and spares the user unnecessary beeps
and typing. Also, now a beep after a <Tab> means just one thing:
there are NO completions.
This fulfills https://savannah.gnu.org/bugs/?58627.
The Spell Checker is no longer among the first twelve items of the
main menu, so the conditional placing of the Go To Line item should
no longer depend on it.
The parameter was referenced in just one place. So, simply check for
the three relevant menus (and unrestricted mode) and be done with it.
This also has the pleasant effect that the menu name is now the first
parameter of do_prompt(), thus clearly indicating what prompt it is,
instead of having an opaque TRUE or FALSE value at the beginning.
Tabbing can list all files in the given directory, but restricted mode
is supposed to prevent the user from any access to the filesystem...
This fixes https://savannah.gnu.org/bugs/?58632.
Bug existed (in this form) since version 1.3.4, commit f7b5d930.
Instead of burdening seven other calls of do_prompt() with a useless
parameter, just check for MGOTODIR in the appropriate place. It also
saves having to pass the parameter down through three more functions.
The single dot serves no purpose, as the user is already there. And
the double dot is reached more easily by typing a second dot first.
And anyway, . and .. are not shown when the user does not type a dot
first, so why show them when the user types a single dot followed by
<Tab><Tab>? Most likely the user wants to see actual dot files, so
just show those.
This fixes https://savannah.gnu.org/bugs/?58619.
Now the help lines can be toggled not only while editing, but also at
the Read (^R), Write (^O), Execute (^T), Search (^W), Replace (M-R),
Goto (^/), and Yesno prompts, and also in the file browser and when
searching for a file name. The help lines cannot be toggled in the
help viewer, nor when searching in a help text, nor in the linter,
as these three things force the help lines to be on.
Furthermore, the 'nohelp' function can be rebound in all relevant
menus (default binding: M-X).
This fulfills https://savannah.gnu.org/bugs/?58471.
The modifier has to come before the foreground color name, separated
by a comma, and after 'bold,' (when present). This has no effect on
terminals that are incapable of cursive text, like the Linux console.
This fulfills https://savannah.gnu.org/bugs/?56525.
The username_tab_completion() function taken from busybox-0.46 was
just a stub, it contained basically nothing. Chris himself wrote
the function in November 2000 (commit be77c611), and rewrote it in
January 2001 (commit 2c2c5f21).
The version of username_tab_completion() in that latter commit looks
somewhat like the version of Vladimir N. Oleynik <dzo@simtreas.ru>
(busybox commit 4e338757). That commit comes three days after Chris'
commit, but Vladimir's original email is from a week earlier:
http://lists.busybox.net/pipermail/busybox/2001-January/035687.html.
So... it is quite possible that Rocco saw that email and suggested
to Chris to rewrite username_tab_completion() using getpwent().
Anyway, the version that was taken from busybox-0.46 was just four
lines and thus not worthy of copyright. The two other functions
that were copied were heavily modified in 2000, so it's better to
add that year in the copyright notice. It is fine to tweak the
notice: it just says that the original authors must be attributed,
not that the exact text must remain unchanged.
Instead of first trying to match things, and then discarding these
matches when the cursor is not at the end of the offered fragment
('buf'), simply don't bother to do any matching in that case.
The variable is set to zero at its declaration, and the second function
is called only when the first is either not called or found nothing --
thus '*num_matches' will still be zero.
Names give a rough idea of what the color will look like, whereas
numerical indexes would not do this at all.
Nine extra colors seems enough. If there were more, no one would be
able to say for sure which is which when two similar colors are used
several rows apart.
This partially fulfills https://savannah.gnu.org/bugs/?56445.
After a year or so, when people have had time to convert their syntaxes
to use 'bold' and/or 'light' instead of 'bright', we can drop this
colorizing, and still later stop recognizing 'bright' altogether.
Having the M-F toggle non-persistent, makes the behavior of ^R
predictable. This allows string binds that load a file to work
correctly independently of when M-F was last pressed.
The -F/--multibuffer command-line option and "set multibuffer"
in a nanorc file determine the default behavior of ^R.
Signed-off-by: Marco Diego Aurélio Mesquita <marcodiegomesquita@gmail.com>
There were no calls of statusbar() or statusline() before curses mode
was entered. But since the previous commit curses mode is entered
even earlier, so...
A background color *can* be bright (in the sense of lighter, shinier).
It's just that the background cannot be bold. Only the typeface used
on the foreground can be bold (if the terminal supports it). So, let
"bright", when used with a background color, simply mean "light".
This makes it possible to avoid the misnamed prefix "bright". It is
misnamed because (on current terminal emulators, when the brightening
of bold is switched off) it just makes the typeface bold, not the color
brighter. The prefix "light" will now only make the color brigther,
and the modifier "bold" will just make the typeface bold (on terminal
emulators, when the brightening of bold is switched off).
On a Linux console, which is not capable of bolding the typeface,
"bold" will brighten the color.
This fully fulfills https://savannah.gnu.org/bugs/?58503.
The prefix "bright" will continue to make the color brighter AND/OR
make the typeface bold (depending on how the terminal is configured
and on its capabilities), but the prefix "light" will just make the
color brighter (if the terminal can do that).
The prefix "bright" continues to do the same as what it did before,
to not change the appearance of user-defined syntaxes. And having
an option to change the meaning of "bright"... would be confusing.
This partially fulfills https://savannah.gnu.org/bugs/?58503.
This way we don't have to probe twice, and the result is accurate also
when nano is reading data from standard input. Standard output should
always be connected to a terminal, as nano is not meant to operate
without a screen.
Benno Schulenberg