Fix regression introduced in 37ffc02422.
(tty_got_winch): don't read FIFO, just check whether it os empty or not.
It can be called many times without lost of info about SIGWINCH raises.
(tty_flush_winch): read from FIFO to make it empty.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
S_Lang: pre2.3.3-8 (55f58798c267d76a1b93d0d916027b71a10ac1ee): removed
static buffers dependent upon SLTT_MAX_SCREEN_ROWS/COLS in favor of
dynamically allocated arrays.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
Use pipe to collect consecutive raised SIGWINCH events.
Then handle multiple events at a time treating them
as a single one.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
How to reproduce the bug:
* compile mc with gpm support
* run the gpm daemon on the console
* login to the machine with ssh from console on another machine
* run mc
At this situation, keyboard response will be sluggish and mc will consume
CPU time when idle. The reason is that mc is trying to reopen the Gpm
connection over and over again and gpm will try to close the connection
over and over again (because the user is not on the local console).
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
Found by GCC 6.2.0.
color-internal.c:186:33: error: format '%X' expects argument of type 'unsigned int', but argument 3 has type 'int' [-Werror=format=]
Signed-off-by: Andreas Mohr <and@gmx.li>
Move calls of disable_mouse() and disable_bracketed_paste()
to main() because init_mouse() and enable_bracketed_paste()
are called here.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
The code that manipulates the ncurses backend into changing
the key combination to generate SIGINT from CTRL-c to CTRL-g does
so by accessing undocumented internal ncurses data structures.
This breaks compilation with netbsd-curses[0], and could also break
when the ncurses author decides to change internal structures in a
future release.
Fix it by using a portable approach that works everywhere using libc
primitives instead.
[0] https://github.com/sabotage-linux/netbsd-curses
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
Introduce -Wswitch-default check.
Some minor cosmetics.
Thanks Andreas Mohr <and at gmx dot li> for original patch.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
mc fails to build when using musl as the libc provider. This is due to
the CTRL() macro not being defined in <termios.h>. We could include
<sys/ttydefaults.h> explicitly but it's easier just to ensure CTRL is
defined.
This patch taken from the Sabotage Linux distro which fixes this. This
patch has also been tested and works with the OpenEmbedded build
system.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
The "topmiddle" and "bottommiddle" characters are defined incorrectly
in many skins. This is because the correct definition wouldn't work
with S-Lang build. The only place these characters are used is the
diffviewer if either the +/- signs or the line numbers are shown,
enabled by pressing S or L. The correct definition of these characters
currently show up as 'v' and 'w' characters.
The fix makes the diffviewer use tty_print_alt_char() as it is being used
in other parts of the code, and hence it fixes the bug.
It also fixes the skin definitions.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
MC already has its own half-ready trick: when pasting with Shift-Insert,
using the X11 extension, the newline ("Enter" as mc calls it) with the
Shift modifier pressed gets converted to a "Return", and in the editor
the Return character inserts a non-indenting newline. This makes pasting
better in terminals not supporting bracketed paste, however, it has some
problems that this commit addresses:
* Shift+newline gets this special treatment, but Ctrl+newline gets
dropped. Hence e.g. when pasting in Gnome-terminal with Ctrl+Shift+V
all the newlines will be missing. This commit adds the same
non-indenting newline behavior to Ctrl+Newline and Ctrl+Shift+Newline.
* The code forgets about Tab that also needs special treatment:
- Most terminals send \e[Z on Shift+Tab, this is not handled by MC
at all, moreover it causes a hang for about a second. This commit
teaches this sequence to MC. This is especially useful when no X11
is available, because there Ctrl+Tab is identical to Tab, so the
backwards tab feature is not available. With this commit Shift+Tab
becomes a backwards tab too on all terminals that emit \e[Z.
- When pasting to the editor, Shift+Tab, Ctrl+Tab and Ctrl+Shift+Tab
should all insert a tab for the same reason mentioned at the newline.
- It would look inconsistent in the keymap files to have logical code
such as "backtab" instead of "shift-tab" and friends, hence get rid
of KEY_BTAB and use KEY_M_SHIFT | '\t' instead.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
HAVE_NCURSESW_CURSES_H included the wrong header.
Partially revert 80c8d58003
for lib/tty/tty-ncurses.h.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
mc may not detect the gpm mouse under X11 on many x terminals when running together with tmux or screen.
On some cases it can even hang.
It's a small patch to the gpm maintainer to improve its checking.
Seems there is a need also to a small change on mc.
Now everything is working perfectly.
Signed-off-by: Slava Zanko <slavazanko@gmail.com>
to drop already received part - there can be more of it
coming over e.g. a serial line.
To prevent interpreting it as a random garbage,
eat and discard all chars that follow.
Small, but non-zero timeout is needed to reconnect
escape sequence split up by a serial line.
Before this change, Ctrl-Alt-Shift-Right_Arrow generates "1;8C"
bogus "input" in MC on my machine; after the change,
nothing is generated.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Slava Zanko <slavazanko@gmail.com>
tail as a garbage input. To reproduce, run "sleep 3" and
hold down Down_Arrow key until sleep runs.
With debugging log enabled, the following can be seen:
entered get_key_code(no_delay:0)
c=tty_lowlevel_getch()=27
push_char(27) !0
c=xgetch_second()=91
push_char(91) !0
2 c=tty_lowlevel_getch()=66
push_char(66)
seq_buffer[0]:27 <---- the saved Down Arrow sequence "ESC [ B"
seq_buffer[1]:91
seq_buffer[2]:66
seq_buffer[3]:0
pending_keys!=NULL. m=-1
d=*pending_keys++=27
d=ALT(*pending_keys++)=ALT(91)=8283
^^^^^^^^^^^^^^^^^^^^^^^ we misinterpret "ESC [ B" as "ESC ["
return correct_key_code(8283)
entered get_key_code(no_delay:0)
pending_keys!=NULL. m=-1
d=*pending_keys++=66
^^^^^^^^^^^^ we think user pressed "B"
return correct_key_code(66)
With this patch, no bogus "input" is generated.
Longer unknown sequences need an additional fix, coming next.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Slava Zanko <slavazanko@gmail.com>
keyboard input: simplify code, no logic changes
This change slightly simplifies and rearranges the code
in get_key_code(), reduces indentation levels there,
adds a few comments. The logic remains the same.
This is a preparatory patch for subsequent changes.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Slava Zanko <slavazanko@gmail.com>
Steps to Reproduce:
1. Run mc in native console (not in X terminal emulator).
2. Press Ctrl+O to switch to subshell.
3. Try select anything with mouse.
Result: mouse does't select anything.
This bug was introduced in 68468a25ac
commit.
Solution: make mouse initialization after initializaton of subshell.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
A followup extension, "SGR 1006" was invented by xterm, to overcome some
of the shortcomings of the previous one. It is becoming as widespread as
the previous one, and is likely to soon overtake it in popularity.
Note that most of the patch is just the removal of the huge complexity
introduced by the previous one. The previous extension didn't have a
unique prefix which made the whole parsing logic extremely complicated.
The new extension does have a unique prefix, so parsing becomes a piece
of cake. The code becomes much cleaner and much easier to maintain.
Signed-off-by: Slava Zanko <slavazanko@gmail.com>
Set defines via CPPFLAGS variable not via CFLAGS one.
Use AM_CPPFLAGS and AM_CFLAGS variables instead of per-target ones.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
Unlike S-Lang, NCurses always wraps long lines and unable trancate that
to pring only visible part of text. Therefore we have to implement our
own addch() wrappers.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
The ancient way of reporting mouse coordinates only supports coordinates
up to 231, so if your terminal is wider (or taller, but that's unlikely),
you cannot use your mouse in the rightmost columns.
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
Since content of keymap file is ASCII-only and case insensitive,
don't use some utf8-manipulation: g_ascii_strcasecmp() is used
instead of str_casecmp().
Signed-off-by: Andrew Borodin <aborodin@vmail.ru>
