magras [Thu, 28 Feb 2019 01:56:01 +0000 (04:56 +0300)]
fix use after free in font caching algorithm
Current font caching algorithm contains a use after free error. A font
removed from `frc` might be still listed in `wx.specbuf`. It will lead
to a crash inside `XftDrawGlyphFontSpec()`.
Steps to reproduce:
$ st -f 'Misc Tamsyn:scalable=false'
$ curl https://www.cl.cam.ac.uk/~mgk25/ucs/examples/UTF-8-demo.txt
Of course, result depends on fonts installed on a system and fontconfig.
In my case, I'm getting consistent segfaults with different fonts.
I replaced a fixed array with a simple unbounded buffer with a constant
growth rate. Cache starts with a capacity of 0, gets increments by 16,
and never shrinks. On my machine after `cat UTF-8-demo.txt` buffer
reaches a capacity of 192. During casual use capacity stays at 0.
Ivan Tham [Tue, 12 Feb 2019 17:41:41 +0000 (18:41 +0100)]
better Input Method Editor (IME) support
Features:
- Allow input methods swap with hotkey (E.g. left ctrl + left shift).
- Over-the-spot pre-editing style, pre-edit data placed over insertion point.
- Restart IME without segmentation fault.
TODO:
- Automatically pickup IME if st started before IME
Paride Legovini [Thu, 10 Jan 2019 12:36:09 +0000 (13:36 +0100)]
Let the user specify CPPFLAGS
This complements the work done in d4928ed, allowing the user to specify
the preprocessor flags with the CPPFLAGS environment variable. This is
useful for example to specify preprocessor macros with -D.
CFLAGS could be used instead, but CPPFLAGS is more correct and is expected
to be honored in some cases. For example, the helper scripts to build
Debian packages make use of CPPFLAGS, but the variable is currently
being ignored unless manually appended to CFLAGS.
Devin J. Pohly [Sat, 24 Feb 2018 22:32:20 +0000 (16:32 -0600)]
Reduce visibility wherever possible
When possible, declare functions/variables static and move struct
definitions out of headers. In order to allow utf8decode to become
internal, use codepoint for DECSCUSR extension directly.
Devin J. Pohly [Sat, 24 Feb 2018 22:16:12 +0000 (16:16 -0600)]
Limit usage of extern to config.h globals
Prefer passing arguments to declaring external global variables. The
only remaining usage of extern is for config.h variables which are
needed in st.c instead of x.c (where it is now included).
Devin J. Pohly [Sat, 24 Feb 2018 21:32:48 +0000 (15:32 -0600)]
Remove x.c dependency on term
The xinit function only needs to the rows/cols, so pass those in rather
than accessing term directly. With a bit of arithmetic, we are able to
avoid the need for term.row and term.col in x2col, y2row, and
xdrawglyphfontspecs as well, completing the removal.
Devin J. Pohly [Sat, 24 Feb 2018 20:58:54 +0000 (14:58 -0600)]
Pull term references out of xdrawcursor
Gradually reducing x.c dependency on Term object. Old and new cursor
glyph/position are passed to xdrawcursor. (There may be an opportunity
to refactor further if we can unify "clear old cursor" and "draw new
cursor" functionality.)
Devin J. Pohly [Sat, 24 Feb 2018 20:53:23 +0000 (14:53 -0600)]
Move win-agnostic parts of draw/drawregion to st.c
Introduces three functions to encapsulate X-specific behavior:
* xdrawline: draws a portion of a single line (used by drawregion)
* xbegindraw: called to prepare for drawing (will be useful for e.g.
Wayland) and returns true if drawing should happen
* xfinishdraw: called to finish drawing (used by draw)
Devin J. Pohly [Fri, 23 Feb 2018 20:16:52 +0000 (14:16 -0600)]
Split mode bits between Term and TermWindow
Moves the mode bits used by x.c from Term to TermWindow, absorbing
UI/input-related mode bits (visible/focused/numlock) along the way.
This is gradually reducing external references to Term. Since
TermWindow is already internal to x.c, we add xsetmode() to allow st to
modify window bits in accordance with escape sequences.
IS_SET() is redefined accordingly (term.mode in st.c, win.mode in x.c).
Devin J. Pohly [Thu, 22 Feb 2018 07:05:12 +0000 (01:05 -0600)]
Move CRLF input processing into ttywrite
This also allows us to remove the crlf field from the Key struct, since
the only difference it made was converting "\r" to "\r\n" (which is now
done automatically in ttywrite). In addition, MODE_CRLF is no longer
referenced from x.c.
Devin J. Pohly [Thu, 22 Feb 2018 05:29:41 +0000 (23:29 -0600)]
Move terminal-related selection logic into st.c
The front-end determines information about mouse clicks and motion, and
the terminal handles the actual selection start/extend/dirty logic by
row and column.
While we're in the neighborhood, we'll also rename getbuttoninfo() to
mousesel() which is, at least, less wrong.
Devin J. Pohly [Thu, 22 Feb 2018 04:28:41 +0000 (22:28 -0600)]
Rely on ttyresize to set tty size
This removes ttynew's dependency on cresize being called first, and then
allows us to absorb the ttyresize call into cresize (which always
precedes it).
Devin J. Pohly [Tue, 17 Oct 2017 20:21:04 +0000 (15:21 -0500)]
Move config.h include from st.c to x.c
config.h includes references to KeySyms and other X stuff. Until we
come up with a cleaner way to separate configuration, it is simpler
(leads to more code removal) to have this here.
Devin J. Pohly [Mon, 16 Oct 2017 01:35:48 +0000 (20:35 -0500)]
Factor out equivalent code from ttyread/ttysend
The echo-to-terminal portions of ttyread and ttysend were actually doing
the same thing. New function twrite() now handles this. The parameter
show_ctrl determines whether control characters are shown as "^A". This
was the only difference between tputc and techo, and techo is now unused
and removed.
Devin J. Pohly [Fri, 13 Oct 2017 03:25:49 +0000 (22:25 -0500)]
Move opt_* into same file as main()/run()
This commit is purely about reducing externs and LOC. If the main and
run functions ever move elsewhere (which will probably make sense
eventually), these should come along with them.
Devin J. Pohly [Wed, 11 Oct 2017 13:47:14 +0000 (08:47 -0500)]
Move key-matching functions into x.c
Modifiers and keysyms are specific to X, and the functions match and
kmap are only used in x.c. Needed to global-ize the key arrays and
lengths from config.h (for now).
Devin J. Pohly [Tue, 10 Oct 2017 15:30:23 +0000 (10:30 -0500)]
Call xsetenv() in main process instead of child
This makes xsetenv internal to x.c, and allows iso14755's external
command to use $WINDOWID instead of having to snprintf it again. (The
same benefit will apply to the externalpipe patch.) The xwinid function
is no longer needed.
Benno Fünfstück [Tue, 26 Dec 2017 15:38:27 +0000 (16:38 +0100)]
Apply ATTR_REVERSE after ATTR_FAINT
An example where the new behaviour makes more sense:
Suppose some text is formatted with ATTR_FAINT for red for the foreground, so it
is rendered in a dark red. In that case, when selected with the mouse, the
intended behaviour is that foreground and background color are swapped: so the
selection should be rendered in dark red and the text in the default background
color.
Before this patch, what happened was that the selection would be in normal red
and the text in the darkened background color, making it almost unreadable.
For an example application that uses the FAINT attribute, try dmesg from
util-linux with color support, it uses FAINT for segfault messages.
this was wrong as pointed out by k0ga:
"STLDFLAGS is about flags to the linker, for example -L
not about -l for that reason it must go before the object list".
Nils Reuße [Wed, 29 Mar 2017 16:34:12 +0000 (18:34 +0200)]
keep some glyph modes for the cursor
st currently does not keep any mode for the cursor that was active
in the underlying glyph (e.g. italic text), the mode is always
ATTR_NULL [1]. At [2] you can find a screenshot that shows the
implications. Other terminals (at least vte-based, such as
XFCE-terminal) keep some modes for the cursor. I find the current
behaviour very disruptive, so here is a patch that keeps a few
(arbitrarily chosen) modes for the cursor.
Roberto E. Vargas Caballero [Wed, 25 Jan 2017 18:17:38 +0000 (19:17 +0100)]
Change default keybindings
CTRL+SHIFT is an impossible combination in the terminal world
(0x20 | x & 0x1F), so it is perfect to be used for internals
shortcuts of terminals, and being a double combination
reduces the prossibility of having comflicts.
Spencer Phippen [Wed, 23 Nov 2016 18:17:59 +0000 (19:17 +0100)]
Fixed 'missing glyph doesn't use fontconfig config substitutions' bug
XftFontMatch does display-specific font configuration (commit 528241a).
Nice. Unfortunately, when we switched from FcFontMatch, we also stopped
storing the post-Fc{Config,Default}Substitute FcPattern for future
lookups. The result is that if a glyph isn't found in the primary font,
secondary font lookups use the original FcPattern, not the configured
one. If you have custom fontconfig rules (like me), this can be
disappointing.
I basically just copied the guts out of XftFontMatch[1] and saved
the intermediate configured FcPattern. Could be related to the bug that
inspired commit 4242027.