ui/term: add line/cell/char handling for terminal pages

This commit introduces libsystemd-ui, a systemd-internal helper library
that will contain all the UI related functionality. It is going to be used
by systemd-welcomed, systemd-consoled, systemd-greeter and systemd-er.
Further use-cases may follow.

For now, this commit only adds terminal-page handling based on lines only.
Follow-up commits will add more functionality.

1 files changed, 1142 insertions, 0 deletions

diff --git a/src/libsystemd-terminal/term-page.c b/src/libsystemd-terminal/term-page.c
new file mode 100644
index 0000000000..bfff3b1719
--- /dev/null
+++ b/src/libsystemd-terminal/term-page.c
@@ -0,0 +1,1142 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+/***
+  This file is part of systemd.
+
+  Copyright (C) 2014 David Herrmann <dh.herrmann@gmail.com>
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 of the License, or
+  (at your option) any later version.
+
+  systemd is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+/*
+ * Terminal Page/Line/Cell/Char Handling
+ * This file implements page handling of a terminal. It is split into pages,
+ * lines, cells and characters. Each object is independent of the next upper
+ * object.
+ *
+ * The Terminal layer keeps each line of a terminal separate and dynamically
+ * allocated. This allows us to move lines from main-screen to history-buffers
+ * very fast. Same is true for scrolling, top/bottom borders and other buffer
+ * operations.
+ *
+ * While lines are dynamically allocated, cells are not. This would be a waste
+ * of memory and causes heavy fragmentation. Furthermore, cells are moved much
+ * less frequently than lines so the performance-penalty is pretty small.
+ * However, to support combining-characters, we have to initialize and cleanup
+ * cells properly and cannot just release the underlying memory. Therefore,
+ * cells are treated as proper objects despite being allocated in arrays.
+ *
+ * Each cell has a set of attributes and a stored character. This is usually a
+ * single Unicode character stored as 32bit UCS-4 char. However, we need to
+ * support Unicode combining-characters, therefore this gets more complicated.
+ * Characters themselves are represented by a "term_char_t" object. It
+ * should be treated as a normal integer and passed by value. The
+ * sorrounding struct is just to hide the internals. A term-char can contain a
+ * base character together with up to 2 combining-chars in a single integer.
+ * Only if you need more combining-chars (very unlikely!) a term-char is a
+ * pointer to an allocated storage. This requires you to always free term-char
+ * objects once no longer used (even though this is a no-op most of the time).
+ * Furthermore, term-char objects are not ref-counted so you must duplicate them
+ * in case you want to store it somewhere and retain a copy yourself. By
+ * convention, all functions that take a term-char object will not duplicate
+ * it but implicitly take ownership of the passed value. It's up to the caller
+ * to duplicate it beforehand, in case it wants to retain a copy.
+ *
+ * If it turns out, that more than 2 comb-chars become common in specific
+ * languages, we can try to optimize this. One idea is to ref-count allocated
+ * characters and store them in a hash-table (like gnome's libvte3 does). This
+ * way we will never have two allocated chars for the same content. Or we can
+ * simply put two uint64_t into a "term_char_t". This will slow down operations
+ * on systems that don't need that many comb-chars, but avoid the dynamic
+ * allocations on others.
+ * Anyhow, until we have proper benchmarks, we will keep the current code. It
+ * seems to compete very well with other solutions so far.
+ */
+
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <wchar.h>
+#include "macro.h"
+#include "term-internal.h"
+#include "util.h"
+
+/* maximum UCS-4 character */
+#define CHAR_UCS4_MAX (0x10ffff)
+/* mask for valid UCS-4 characters (21bit) */
+#define CHAR_UCS4_MASK (0x1fffff)
+/* UCS-4 replacement character */
+#define CHAR_UCS4_REPLACEMENT (0xfffd)
+
+/* real storage behind "term_char_t" in case it's not packed */
+typedef struct term_character {
+        uint8_t n;
+        uint32_t codepoints[];
+} term_character;
+
+/*
+ * char_pack() takes 3 UCS-4 values and packs them into a term_char_t object.
+ * Note that UCS-4 chars only take 21 bits, so we still have the LSB as marker.
+ * We set it to 1 so others can distinguish it from pointers.
+ */
+static inline term_char_t char_pack(uint32_t v1, uint32_t v2, uint32_t v3) {
+        uint64_t packed, u1, u2, u3;
+
+        u1 = v1;
+        u2 = v2;
+        u3 = v3;
+
+        packed = 0x01;
+        packed |= (u1 & (uint64_t)CHAR_UCS4_MASK) << 43;
+        packed |= (u2 & (uint64_t)CHAR_UCS4_MASK) << 22;
+        packed |= (u3 & (uint64_t)CHAR_UCS4_MASK) <<  1;
+
+        return TERM_CHAR_INIT(packed);
+}
+
+#define char_pack1(_v1) char_pack2((_v1), CHAR_UCS4_MAX + 1)
+#define char_pack2(_v1, _v2) char_pack3((_v1), (_v2), CHAR_UCS4_MAX + 1)
+#define char_pack3(_v1, _v2, _v3) char_pack((_v1), (_v2), (_v3))
+
+/*
+ * char_unpack() is the inverse of char_pack(). It extracts the 3 stored UCS-4
+ * characters and returns them. Note that this does not validate the passed
+ * term_char_t. That's the responsibility of the caller.
+ * This returns the number of characters actually packed. This obviously is a
+ * number between 0 and 3 (inclusive).
+ */
+static inline uint8_t char_unpack(term_char_t packed, uint32_t *out_v1, uint32_t *out_v2, uint32_t *out_v3) {
+        uint32_t v1, v2, v3;
+
+        v1 = (packed._value >> 43) & (uint64_t)CHAR_UCS4_MASK;
+        v2 = (packed._value >> 22) & (uint64_t)CHAR_UCS4_MASK;
+        v3 = (packed._value >>  1) & (uint64_t)CHAR_UCS4_MASK;
+
+        if (out_v1)
+                *out_v1 = v1;
+        if (out_v2)
+                *out_v2 = v2;
+        if (out_v3)
+                *out_v3 = v3;
+
+        return (v1 > CHAR_UCS4_MAX) ? 0 :
+              ((v2 > CHAR_UCS4_MAX) ? 1 :
+              ((v3 > CHAR_UCS4_MAX) ? 2 :
+                                      3));
+}
+
+/* cast a term_char_t to a term_character* */
+static inline term_character *char_to_ptr(term_char_t ch) {
+        return (term_character*)(unsigned long)ch._value;
+}
+
+/* cast a term_character* to a term_char_t */
+static inline term_char_t char_from_ptr(term_character *c) {
+        return TERM_CHAR_INIT((unsigned long)c);
+}
+
+/*
+ * char_alloc() allocates a properly aligned term_character object and returns
+ * a pointer to it. NULL is returned on allocation errors. The object will have
+ * enough room for @n following UCS-4 chars.
+ * Note that we allocate (n+1) characters and set the last one to 0 in case
+ * anyone prints this string for debugging.
+ */
+static term_character *char_alloc(uint8_t n) {
+        term_character *c;
+        int r;
+
+        r = posix_memalign((void**)&c,
+                           MAX(sizeof(void*), (size_t)2),
+                           sizeof(*c) + sizeof(*c->codepoints) * (n + 1));
+        if (r)
+                return NULL;
+
+        c->n = n;
+        c->codepoints[n] = 0;
+
+        return c;
+}
+
+/*
+ * char_free() frees the memory allocated via char_alloc(). It is safe to call
+ * this on any term_char_t, only allocated characters are freed.
+ */
+static inline void char_free(term_char_t ch) {
+        if (term_char_is_allocated(ch))
+                free(char_to_ptr(ch));
+}
+
+/*
+ * This appends @append_ucs4 to the existing character @base and returns
+ * it as a new character. In case that's not possible, @base is returned. The
+ * caller can use term_char_same() to test whether the returned character was
+ * freshly allocated or not.
+ */
+static term_char_t char_build(term_char_t base, uint32_t append_ucs4) {
+        /* soft-limit for combining-chars; hard-limit is currently 255 */
+        const size_t climit = 64;
+        term_character *c;
+        uint32_t buf[3], *t;
+        uint8_t n;
+
+        /* ignore invalid UCS-4 */
+        if (append_ucs4 > CHAR_UCS4_MAX)
+                return base;
+
+        if (term_char_is_null(base)) {
+                return char_pack1(append_ucs4);
+        } else if (!term_char_is_allocated(base)) {
+                /* unpack and try extending the packed character */
+                n = char_unpack(base, &buf[0], &buf[1], &buf[2]);
+
+                switch (n) {
+                case 0:
+                        return char_pack1(append_ucs4);
+                case 1:
+                        if (climit < 2)
+                                return base;
+
+                        return char_pack2(buf[0], append_ucs4);
+                case 2:
+                        if (climit < 3)
+                                return base;
+
+                        return char_pack3(buf[0], buf[1], append_ucs4);
+                default:
+                        /* fallthrough */
+                        break;
+                }
+
+                /* already fully packed, we need to allocate a new one */
+                t = buf;
+        } else {
+                /* already an allocated type, we need to allocate a new one */
+                c = char_to_ptr(base);
+                t = c->codepoints;
+                n = c->n;
+        }
+
+        /* bail out if soft-limit is reached */
+        if (n >= climit)
+                return base;
+
+        /* allocate new char */
+        c = char_alloc(n + 1);
+        if (!c)
+                return base;
+
+        memcpy(c->codepoints, t, sizeof(*t) * n);
+        c->codepoints[n] = append_ucs4;
+
+        return char_from_ptr(c);
+}
+
+/**
+ * term_char_set() - Reset character to a single UCS-4 character
+ * @previous: term-char to reset
+ * @append_ucs4: UCS-4 char to set
+ *
+ * This frees all resources in @previous and re-initializes it to @append_ucs4.
+ * The new char is returned.
+ *
+ * Usually, this is used like this:
+ *   obj->ch = term_char_set(obj->ch, ucs4);
+ *
+ * Returns: The previous character reset to @append_ucs4.
+ */
+term_char_t term_char_set(term_char_t previous, uint32_t append_ucs4) {
+        char_free(previous);
+        return char_build(TERM_CHAR_NULL, append_ucs4);
+}
+
+/**
+ * term_char_merge() - Merge UCS-4 char at the end of an existing char
+ * @base: existing term-char
+ * @append_ucs4: UCS-4 character to append
+ *
+ * This appends @append_ucs4 to @base and returns the result. @base is
+ * invalidated by this function and must no longer be used. The returned value
+ * replaces the old one.
+ *
+ * Usually, this is used like this:
+ *   obj->ch = term_char_merge(obj->ch, ucs4);
+ *
+ * Returns: The new merged character.
+ */
+term_char_t term_char_merge(term_char_t base, uint32_t append_ucs4) {
+        term_char_t ch;
+
+        ch = char_build(base, append_ucs4);
+        if (!term_char_same(ch, base))
+                term_char_free(base);
+
+        return ch;
+}
+
+/**
+ * term_char_dup() - Duplicate character
+ * @ch: character to duplicate
+ *
+ * This duplicates a term-character. In case the character is not allocated,
+ * nothing is done. Otherwise, the underlying memory is copied and returned. You
+ * need to call term_char_free() on the returned character to release it again.
+ * On allocation errors, a replacement character is returned. Therefore, the
+ * caller can safely assume that this function always succeeds.
+ *
+ * Returns: The duplicated term-character.
+ */
+term_char_t term_char_dup(term_char_t ch) {
+        term_character *c, *newc;
+
+        if (!term_char_is_allocated(ch))
+                return ch;
+
+        c = char_to_ptr(ch);
+        newc = char_alloc(c->n);
+        if (!newc)
+                return char_pack1(CHAR_UCS4_REPLACEMENT);
+
+        memcpy(newc->codepoints, c->codepoints, sizeof(*c->codepoints) * c->n);
+        return char_from_ptr(newc);
+}
+
+/**
+ * term_char_dup_append() - Duplicate tsm-char with UCS-4 character appended
+ * @base: existing term-char
+ * @append_ucs4: UCS-4 character to append
+ *
+ * This is similar to term_char_merge(), but it returns a separately allocated
+ * character. That is, @base will stay valid after this returns and is not
+ * touched. In case the append-operation fails, @base is duplicated and
+ * returned. That is, the returned char is always independent of @base.
+ *
+ * Returns: Newly allocated character with @append_ucs4 appended to @base.
+ */
+term_char_t term_char_dup_append(term_char_t base, uint32_t append_ucs4) {
+        term_char_t ch;
+
+        ch = char_build(base, append_ucs4);
+        if (term_char_same(ch, base))
+                ch = term_char_dup(base);
+
+        return ch;
+}
+
+/**
+ * term_char_resolve() - Retrieve the UCS-4 string for a term-char
+ * @ch: character to resolve
+ * @s: storage for size of string or NULL
+ * @b: storage for string or NULL
+ *
+ * This takes a term-character and returns the UCS-4 string associated with it.
+ * In case @ch is not allocated, the string is stored in @b (in case @b is NULL
+ * static storage is used). Otherwise, a pointer to the allocated storage is
+ * returned.
+ *
+ * The returned string is only valid as long as @ch and @b are valid. The string
+ * is zero-terminated and can safely be printed via long-character printf().
+ * The length of the string excluding the zero-character is returned in @s.
+ *
+ * This never returns NULL. Even if the size is 0, this points to a buffer of at
+ * least a zero-terminator.
+ *
+ * Returns: The UCS-4 string-representation of @ch, and its size in @s.
+ */
+const uint32_t *term_char_resolve(term_char_t ch, size_t *s, term_charbuf_t *b) {
+        static term_charbuf_t static_b;
+        term_character *c;
+        uint32_t *cache;
+        size_t len;
+
+        if (b)
+                cache = b->buf;
+        else
+                cache = static_b.buf;
+
+        if (term_char_is_null(ch)) {
+                len = 0;
+                cache[0] = 0;
+        } else if (term_char_is_allocated(ch)) {
+                c = char_to_ptr(ch);
+                len = c->n;
+                cache = c->codepoints;
+        } else {
+                len = char_unpack(ch, &cache[0], &cache[1], &cache[2]);
+                cache[len] = 0;
+        }
+
+        if (s)
+                *s = len;
+
+        return cache;
+}
+
+/**
+ * term_char_lookup_width() - Lookup cell-width of a character
+ * @ch: character to return cell-width for
+ *
+ * This is an equivalent of wcwidth() for term_char_t. It can deal directly
+ * with UCS-4 and combining-characters and avoids the mess that is wchar_t and
+ * locale handling.
+ *
+ * Returns: 0 for unprintable characters, >0 for everything else.
+ */
+unsigned int term_char_lookup_width(term_char_t ch) {
+        term_charbuf_t b;
+        const uint32_t *str;
+        unsigned int max;
+        size_t i, len;
+        int r;
+
+        max = 0;
+        str = term_char_resolve(ch, &len, &b);
+
+        for (i = 0; i < len; ++i) {
+                /*
+                 * Oh god, C99 locale handling strikes again: wcwidth() expects
+                 * wchar_t, but there is no way for us to know the
+                 * internal encoding of wchar_t. Moreover, it is nearly
+                 * impossible to convert UCS-4 into wchar_t (except for iconv,
+                 * which is way too much overhead).
+                 * Therefore, we use our own copy of wcwidth(). Lets just hope
+                 * that glibc will one day export it's internal UCS-4 and UTF-8
+                 * helpers for direct use.
+                 */
+                assert_cc(sizeof(wchar_t) >= 4);
+                r = mk_wcwidth((wchar_t)str[i]);
+                if (r > 0 && (unsigned int)r > max)
+                        max = r;
+        }
+
+        return max;
+}
+
+/**
+ * term_cell_init() - Initialize a new cell
+ * @cell: cell to initialize
+ * @ch: character to set on the cell or TERM_CHAR_NULL
+ * @cwidth: character width of @ch
+ * @attr: attributes to set on the cell or NULL
+ * @age: age to set on the cell or TERM_AGE_NULL
+ *
+ * This initializes a new cell. The backing-memory of the cell must be allocated
+ * by the caller beforehand. The caller is responsible to destroy the cell via
+ * term_cell_destroy() before freeing the backing-memory.
+ *
+ * It is safe (and supported!) to use:
+ *   zero(*c);
+ * instead of:
+ *   term_cell_init(c, TERM_CHAR_NULL, NULL, TERM_AGE_NULL);
+ *
+ * Note that this call takes ownership of @ch. If you want to use it yourself
+ * after this call, you need to duplicate it before calling this.
+ */
+static void term_cell_init(term_cell *cell, term_char_t ch, unsigned int cwidth, const term_attr *attr, term_age_t age) {
+        assert(cell);
+
+        cell->ch = ch;
+        cell->cwidth = cwidth;
+        cell->age = age;
+
+        if (attr)
+                memcpy(&cell->attr, attr, sizeof(*attr));
+        else
+                zero(cell->attr);
+}
+
+/**
+ * term_cell_destroy() - Destroy previously initialized cell
+ * @cell: cell to destroy or NULL
+ *
+ * This releases all resources associated with a cell. The backing memory is
+ * kept as-is. It's the responsibility of the caller to manage it.
+ *
+ * You must not call any other cell operations on this cell after this call
+ * returns. You must re-initialize the cell via term_cell_init() before you can
+ * use it again.
+ *
+ * If @cell is NULL, this is a no-op.
+ */
+static void term_cell_destroy(term_cell *cell) {
+        if (!cell)
+                return;
+
+        term_char_free(cell->ch);
+}
+
+/**
+ * term_cell_set() - Change contents of a cell
+ * @cell: cell to modify
+ * @ch: character to set on the cell or cell->ch
+ * @cwidth: character width of @ch or cell->cwidth
+ * @attr: attributes to set on the cell or NULL
+ * @age: age to set on the cell or cell->age
+ *
+ * This changes the contents of a cell. It can be used to change the character,
+ * attributes and age. To keep the current character, pass cell->ch as @ch. To
+ * reset the current attributes, pass NULL. To keep the current age, pass
+ * cell->age.
+ *
+ * This call takes ownership of @ch. You need to duplicate it first, in case you
+ * want to use it for your own purposes after this call.
+ *
+ * The cell must have been initialized properly before calling this. See
+ * term_cell_init().
+ */
+static void term_cell_set(term_cell *cell, term_char_t ch, unsigned int cwidth, const term_attr *attr, term_age_t age) {
+        assert(cell);
+
+        if (!term_char_same(ch, cell->ch)) {
+                term_char_free(cell->ch);
+                cell->ch = ch;
+        }
+
+        cell->cwidth = cwidth;
+        cell->age = age;
+
+        if (attr)
+                memcpy(&cell->attr, attr, sizeof(*attr));
+        else
+                zero(cell->attr);
+}
+
+/**
+ * term_cell_append() - Append a combining-char to a cell
+ * @cell: cell to modify
+ * @ucs4: UCS-4 character to append to the cell
+ * @age: new age to set on the cell or cell->age
+ *
+ * This appends a combining-character to a cell. No validation of the UCS-4
+ * character is done, so this can be used to append any character. Additionally,
+ * this can update the age of the cell.
+ *
+ * The cell must have been initialized properly before calling this. See
+ * term_cell_init().
+ */
+static void term_cell_append(term_cell *cell, uint32_t ucs4, term_age_t age) {
+        assert(cell);
+
+        cell->ch = term_char_merge(cell->ch, ucs4);
+        cell->age = age;
+}
+
+/**
+ * term_cell_init_n() - Initialize an array of cells
+ * @cells: pointer to an array of cells to initialize
+ * @n: number of cells
+ * @attr: attributes to set on all cells or NULL
+ * @age: age to set on all cells
+ *
+ * This is the same as term_cell_init() but initializes an array of cells.
+ * Furthermore, this always sets the character to TERM_CHAR_NULL.
+ * If you want to set a specific characters on all cells, you need to hard-code
+ * this loop and duplicate the character for each cell.
+ */
+static void term_cell_init_n(term_cell *cells, unsigned int n, const term_attr *attr, term_age_t age) {
+        for ( ; n > 0; --n, ++cells)
+                term_cell_init(cells, TERM_CHAR_NULL, 0, attr, age);
+}
+
+/**
+ * term_cell_destroy_n() - Destroy an array of cells
+ * @cells: pointer to an array of cells to destroy
+ * @n: number of cells
+ *
+ * This is the same as term_cell_destroy() but destroys an array of cells.
+ */
+static void term_cell_destroy_n(term_cell *cells, unsigned int n) {
+        for ( ; n > 0; --n, ++cells)
+                term_cell_destroy(cells);
+}
+
+/**
+ * term_cell_clear_n() - Clear contents of an array of cells
+ * @cells: pointer to an array of cells to modify
+ * @n: number of cells
+ * @attr: attributes to set on all cells or NULL
+ * @age: age to set on all cells
+ *
+ * This is the same as term_cell_set() but operates on an array of cells. Note
+ * that all characters are always set to TERM_CHAR_NULL, unlike term_cell_set()
+ * which takes the character as argument.
+ * If you want to set a specific characters on all cells, you need to hard-code
+ * this loop and duplicate the character for each cell.
+ */
+static void term_cell_clear_n(term_cell *cells, unsigned int n, const term_attr *attr, term_age_t age) {
+        for ( ; n > 0; --n, ++cells)
+                term_cell_set(cells, TERM_CHAR_NULL, 0, attr, age);
+}
+
+/**
+ * term_line_new() - Allocate a new line
+ * @out: place to store pointer to new line
+ *
+ * This allocates and initialized a new line. The line is unlinked and
+ * independent of any page. It can be used for any purpose. The initial
+ * cell-count is set to 0.
+ *
+ * The line has to be freed via term_line_free() once it's no longer needed.
+ *
+ * Returns: 0 on success, negative error code on failure.
+ */
+int term_line_new(term_line **out) {
+        _term_line_free_ term_line *line = NULL;
+
+        assert_return(out, -EINVAL);
+
+        line = new0(term_line, 1);
+        if (!line)
+                return -ENOMEM;
+
+        *out = line;
+        line = NULL;
+        return 0;
+}
+
+/**
+ * term_line_free() - Free a line
+ * @line: line to free or NULL
+ *
+ * This frees a line that was previously allocated via term_line_free(). All its
+ * cells are released, too.
+ *
+ * If @line is NULL, this is a no-op.
+ */
+term_line *term_line_free(term_line *line) {
+        if (!line)
+                return NULL;
+
+        term_cell_destroy_n(line->cells, line->n_cells);
+        free(line->cells);
+        free(line);
+
+        return NULL;
+}
+
+/**
+ * term_line_reserve() - Pre-allocate cells for a line
+ * @line: line to pre-allocate cells for
+ * @width: numbers of cells the line shall have pre-allocated
+ * @attr: attribute for all allocated cells or NULL
+ * @age: current age for all modifications
+ * @protect_width: width to protect from erasure
+ *
+ * This pre-allocates cells for this line. Please note that @width is the number
+ * of cells the line is guaranteed to have allocated after this call returns.
+ * It's not the number of cells that are added, neither is it the new width of
+ * the line.
+ *
+ * This function never frees memory. That is, reducing the line-width will
+ * always succeed, same is true for increasing the width to a previously set
+ * width.
+ *
+ * @attr and @age are used to initialize new cells. Additionally, any
+ * existing cell outside of the protected area specified by @protect_width are
+ * cleared and reset with @attr and @age.
+ *
+ * Returns: 0 on success, negative error code on failure.
+ */
+int term_line_reserve(term_line *line, unsigned int width, const term_attr *attr, term_age_t age, unsigned int protect_width) {
+        unsigned int min_width;
+        term_cell *t;
+
+        assert_return(line, -EINVAL);
+
+        /* reset existing cells if required */
+        min_width = MIN(line->n_cells, width);
+        if (min_width > protect_width)
+                term_cell_clear_n(line->cells + protect_width,
+                                  min_width - protect_width,
+                                  attr,
+                                  age);
+
+        /* allocate new cells if required */
+
+        if (width > line->n_cells) {
+                t = realloc_multiply(line->cells, sizeof(*t), width);
+                if (!t)
+                        return -ENOMEM;
+
+                if (!attr && !age)
+                        memzero(t + line->n_cells,
+                                sizeof(*t) * (width - line->n_cells));
+                else
+                        term_cell_init_n(t + line->n_cells,
+                                         width - line->n_cells,
+                                         attr,
+                                         age);
+
+                line->cells = t;
+                line->n_cells = width;
+        }
+
+        line->fill = MIN(line->fill, protect_width);
+
+        return 0;
+}
+
+/**
+ * term_line_set_width() - Change width of a line
+ * @line: line to modify
+ * @width: new width
+ *
+ * This changes the actual width of a line. It is the caller's responsibility
+ * to use term_line_reserve() to make sure enough space is allocated. If @width
+ * is greater than the allocated size, it is cropped.
+ *
+ * This does not modify any cells. Use term_line_reserve() or term_line_erase()
+ * to clear any newly added cells.
+ *
+ * NOTE: The fill state is cropped at line->width. Therefore, if you increase
+ *       the line-width afterwards, but there is a multi-cell character at the
+ *       end of the line that got cropped, then the fill-state will _not_ be
+ *       adjusted.
+ *       This means, the fill-state always includes the cells up to the start
+ *       of the right-most character, but it might or might not cover it until
+ *       its end. This should be totally fine, though. You should never access
+ *       multi-cell tails directly, anyway.
+ */
+void term_line_set_width(term_line *line, unsigned int width) {
+        assert(line);
+
+        if (width > line->n_cells)
+                width = line->n_cells;
+
+        line->width = width;
+        line->fill = MIN(line->fill, width);
+}
+
+/**
+ * line_insert() - Insert characters and move existing cells to the right
+ * @from: position to insert cells at
+ * @num: number of cells to insert
+ * @head_char: character that is set on the first cell
+ * @head_cwidth: character-length of @head_char
+ * @attr: attribute for all inserted cells or NULL
+ * @age: current age for all modifications
+ *
+ * The INSERT operation (or writes with INSERT_MODE) writes data at a specific
+ * position on a line and shifts the existing cells to the right. Cells that are
+ * moved beyond the right hand border are discarded.
+ *
+ * This helper contains the actual INSERT implementation which is independent of
+ * the data written. It works on cells, not on characters. The first cell is set
+ * to @head_char, all others are reset to TERM_CHAR_NULL. See each caller for a
+ * more detailed description.
+ */
+static inline void line_insert(term_line *line, unsigned int from, unsigned int num, term_char_t head_char, unsigned int head_cwidth, const term_attr *attr, term_age_t age) {
+        unsigned int i, rem, move;
+
+        if (from >= line->width)
+                return;
+        if (from + num < from || from + num > line->width)
+                num = line->width - from;
+        if (!num)
+                return;
+
+        move = line->width - from - num;
+        rem = MIN(num, move);
+
+        if (rem > 0) {
+                /*
+                 * Make room for @num cells; shift cells to the right if
+                 * required. @rem is the number of remaining cells that we will
+                 * knock off on the right and overwrite during the right shift.
+                 *
+                 * For INSERT_MODE, @num/@rem are usually 1 or 2, @move is 50%
+                 * of the line on average. Therefore, the actual move is quite
+                 * heavy and we can safely invalidate cells manually instead of
+                 * the whole line.
+                 * However, for INSERT operations, any parameters are
+                 * possible. But we cannot place any assumption on its usage
+                 * across applications, so we just handle it the same as
+                 * INSERT_MODE and do per-cell invalidation.
+                 */
+
+                /* destroy cells that are knocked off on the right */
+                term_cell_destroy_n(line->cells + line->width - rem, rem);
+
+                /* move remaining bulk of cells */
+                memmove(line->cells + from + num,
+                        line->cells + from,
+                        sizeof(*line->cells) * move);
+
+                /* invalidate cells */
+                for (i = 0; i < move; ++i)
+                        line->cells[from + num + i].age = age;
+
+                /* initialize fresh head-cell */
+                term_cell_init(line->cells + from,
+                               head_char,
+                               head_cwidth,
+                               attr,
+                               age);
+
+                /* initialize fresh tail-cells */
+                term_cell_init_n(line->cells + from + 1,
+                                 num - 1,
+                                 attr,
+                                 age);
+
+                /* adjust fill-state */
+                DISABLE_WARNING_SHADOW;
+                line->fill = MIN(line->width,
+                                 MAX(line->fill + num,
+                                     from + num));
+                REENABLE_WARNING;
+        } else {
+                /* modify head-cell */
+                term_cell_set(line->cells + from,
+                              head_char,
+                              head_cwidth,
+                              attr,
+                              age);
+
+                /* reset tail-cells */
+                term_cell_clear_n(line->cells + from + 1,
+                                  num - 1,
+                                  attr,
+                                  age);
+
+                /* adjust fill-state */
+                line->fill = line->width;
+        }
+}
+
+/**
+ * term_line_write() - Write to a single, specific cell
+ * @line: line to write to
+ * @pos_x: x-position of cell in @line to write to
+ * @ch: character to write to the cell
+ * @cwidth: character width of @ch
+ * @attr: attributes to set on the cell or NULL
+ * @age: current age for all modifications
+ * @insert_mode: true if INSERT-MODE is enabled
+ *
+ * This writes to a specific cell in a line. The cell is addressed by its
+ * X-position @pos_x. If that cell does not exist, this is a no-op.
+ *
+ * @ch and @attr are set on this cell.
+ *
+ * If @insert_mode is true, this inserts the character instead of overwriting
+ * existing data (existing data is now moved to the right before writing).
+ *
+ * This function is the low-level handler of normal writes to a terminal.
+ */
+void term_line_write(term_line *line, unsigned int pos_x, term_char_t ch, unsigned int cwidth, const term_attr *attr, term_age_t age, bool insert_mode) {
+        unsigned int len;
+
+        assert(line);
+
+        if (pos_x >= line->width)
+                return;
+
+        len = MAX(1U, cwidth);
+        if (pos_x + len < pos_x || pos_x + len > line->width)
+                len = line->width - pos_x;
+        if (!len)
+                return;
+
+        if (insert_mode) {
+                /* Use line_insert() to insert the character-head and fill
+                 * the remains with NULLs. */
+                line_insert(line, pos_x, len, ch, cwidth, attr, age);
+        } else {
+                /* modify head-cell */
+                term_cell_set(line->cells + pos_x, ch, cwidth, attr, age);
+
+                /* reset tail-cells */
+                term_cell_clear_n(line->cells + pos_x + 1,
+                                  len - 1,
+                                  attr,
+                                  age);
+
+                /* adjust fill-state */
+                DISABLE_WARNING_SHADOW;
+                line->fill = MIN(line->width,
+                                 MAX(line->fill,
+                                     pos_x + len));
+                REENABLE_WARNING;
+        }
+}
+
+/**
+ * term_line_insert() - Insert empty cells
+ * @line: line to insert empty cells into
+ * @from: x-position where to insert cells
+ * @num: number of cells to insert
+ * @attr: attributes to set on the cells or NULL
+ * @age: current age for all modifications
+ *
+ * This inserts @num empty cells at position @from in line @line. All existing
+ * cells to the right are shifted to make room for the new cells. Cells that get
+ * pushed beyond the right hand border are discarded.
+ */
+void term_line_insert(term_line *line, unsigned int from, unsigned int num, const term_attr *attr, term_age_t age) {
+        /* use line_insert() to insert @num empty cells */
+        return line_insert(line, from, num, TERM_CHAR_NULL, 0, attr, age);
+}
+
+/**
+ * term_line_delete() - Delete cells from line
+ * @line: line to delete cells from
+ * @from: position to delete cells at
+ * @num: number of cells to delete
+ * @attr: attributes to set on any new cells
+ * @age: current age for all modifications
+ *
+ * Delete cells from a line. All cells to the right of the deleted cells are
+ * shifted to the left to fill the empty space. New cells appearing on the right
+ * hand border are cleared and initialized with @attr.
+ */
+void term_line_delete(term_line *line, unsigned int from, unsigned int num, const term_attr *attr, term_age_t age) {
+        unsigned int rem, move, i;
+
+        assert(line);
+
+        if (from >= line->width)
+                return;
+        if (from + num < from || from + num > line->width)
+                num = line->width - from;
+        if (!num)
+                return;
+
+        /* destroy and move as many upfront as possible */
+        move = line->width - from - num;
+        rem = MIN(num, move);
+        if (rem > 0) {
+                /* destroy to be removed cells */
+                term_cell_destroy_n(line->cells + from, rem);
+
+                /* move tail upfront */
+                memmove(line->cells + from,
+                        line->cells + from + num,
+                        sizeof(*line->cells) * move);
+
+                /* invalidate copied cells */
+                for (i = 0; i < move; ++i)
+                        line->cells[from + i].age = age;
+
+                /* initialize tail that was moved away */
+                term_cell_init_n(line->cells + line->width - rem,
+                                 rem,
+                                 attr,
+                                 age);
+
+                /* reset remaining cells in case the move was too small */
+                if (num > move)
+                        term_cell_clear_n(line->cells + from + move,
+                                          num - move,
+                                          attr,
+                                          age);
+        } else {
+                /* reset cells */
+                term_cell_clear_n(line->cells + from,
+                                  num,
+                                  attr,
+                                  age);
+        }
+
+        /* adjust fill-state */
+        if (from + num < line->fill)
+                line->fill -= num;
+        else if (from < line->fill)
+                line->fill = from;
+}
+
+/**
+ * term_line_append_combchar() - Append combining char to existing cell
+ * @line: line to modify
+ * @pos_x: position of cell to append combining char to
+ * @ucs4: combining character to append
+ * @age: current age for all modifications
+ *
+ * Unicode allows trailing combining characters, which belong to the
+ * char in front of them. The caller is responsible of detecting
+ * combining characters and calling term_line_append_combchar() instead of
+ * term_line_write(). This simply appends the char to the correct cell then.
+ * If the cell is not in the visible area, this call is skipped.
+ *
+ * Note that control-sequences are not 100% compatible with combining
+ * characters as they require delayed parsing. However, we must handle
+ * control-sequences immediately. Therefore, there might be trailing
+ * combining chars that should be discarded by the parser.
+ * However, to prevent programming errors, we're also being pedantic
+ * here and discard weirdly placed combining chars. This prevents
+ * situations were invalid content is parsed into the terminal and you
+ * might end up with cells containing only combining chars.
+ *
+ * Long story short: To get combining-characters working with old-fashioned
+ * terminal-emulation, we parse them exclusively for direct cell-writes. Other
+ * combining-characters are usually simply discarded and ignored.
+ */
+void term_line_append_combchar(term_line *line, unsigned int pos_x, uint32_t ucs4, term_age_t age) {
+        assert(line);
+
+        if (pos_x >= line->width)
+                return;
+
+        /* Unused cell? Skip appending any combining chars then. */
+        if (term_char_is_null(line->cells[pos_x].ch))
+                return;
+
+        term_cell_append(line->cells + pos_x, ucs4, age);
+}
+
+/**
+ * term_line_erase() - Erase parts of a line
+ * @line: line to modify
+ * @from: position to start the erase
+ * @num: number of cells to erase
+ * @attr: attributes to initialize erased cells with
+ * @age: current age for all modifications
+ * @keep_protected: true if protected cells should be kept
+ *
+ * This is the standard erase operation. It clears all cells in the targetted
+ * area and re-initializes them. Cells to the right are not shifted left, you
+ * must use DELETE to achieve that. Cells outside the visible area are skipped.
+ *
+ * If @keep_protected is true, protected cells will not be erased.
+ */
+void term_line_erase(term_line *line, unsigned int from, unsigned int num, const term_attr *attr, term_age_t age, bool keep_protected) {
+        term_cell *cell;
+        unsigned int i, last_protected;
+
+        assert(line);
+
+        if (from >= line->width)
+                return;
+        if (from + num < from || from + num > line->width)
+                num = line->width - from;
+        if (!num)
+                return;
+
+        last_protected = 0;
+        for (i = 0; i < num; ++i) {
+                cell = line->cells + from + i;
+                if (keep_protected && cell->attr.protect) {
+                        /* only count protected-cells inside the fill-region */
+                        if (from + i < line->fill)
+                                last_protected = from + i;
+
+                        continue;
+                }
+
+                term_cell_set(cell, TERM_CHAR_NULL, 0, attr, age);
+        }
+
+        /* Adjust fill-state. This is a bit tricks, we can only adjust it in
+         * case the erase-region starts inside the fill-region and ends at the
+         * tail or beyond the fill-region. Otherwise, the current fill-state
+         * stays as it was.
+         * Furthermore, we must account for protected cells. The loop above
+         * ensures that protected-cells are only accounted for if they're
+         * inside the fill-region. */
+        if (from < line->fill && from + num >= line->fill)
+                line->fill = MAX(from, last_protected);
+}
+
+/**
+ * term_line_reset() - Reset a line
+ * @line: line to reset
+ * @attr: attributes to initialize all cells with
+ * @age: current age for all modifications
+ *
+ * This resets all visible cells of a line and sets their attributes and ages
+ * to @attr and @age. This is equivalent to erasing a whole line via
+ * term_line_erase().
+ */
+void term_line_reset(term_line *line, const term_attr *attr, term_age_t age) {
+        assert(line);
+
+        return term_line_erase(line, 0, line->width, attr, age, 0);
+}
+
+/**
+ * term_line_link() - Link line in front of a list
+ * @line: line to link
+ * @first: member pointing to first entry
+ * @last: member pointing to last entry
+ *
+ * This links a line into a list of lines. The line is inserted at the front and
+ * must not be linked, yet. See the TERM_LINE_LINK() macro for an easier usage of
+ * this.
+ */
+void term_line_link(term_line *line, term_line **first, term_line **last) {
+        assert(line);
+        assert(first);
+        assert(last);
+        assert(!line->lines_prev);
+        assert(!line->lines_next);
+
+        line->lines_prev = NULL;
+        line->lines_next = *first;
+        if (*first)
+                (*first)->lines_prev = line;
+        else
+                *last = line;
+        *first = line;
+}
+
+/**
+ * term_line_link_tail() - Link line at tail of a list
+ * @line: line to link
+ * @first: member pointing to first entry
+ * @last: member pointing to last entry
+ *
+ * Same as term_line_link() but links the line at the tail.
+ */
+void term_line_link_tail(term_line *line, term_line **first, term_line **last) {
+        assert(line);
+        assert(first);
+        assert(last);
+        assert(!line->lines_prev);
+        assert(!line->lines_next);
+
+        line->lines_next = NULL;
+        line->lines_prev = *last;
+        if (*last)
+                (*last)->lines_next = line;
+        else
+                *first = line;
+        *last = line;
+}
+
+/**
+ * term_line_unlink() - Unlink line from a list
+ * @line: line to unlink
+ * @first: member pointing to first entry
+ * @last: member pointing to last entry
+ *
+ * This unlinks a previously linked line. See TERM_LINE_UNLINK() for an easier to
+ * use macro.
+ */
+void term_line_unlink(term_line *line, term_line **first, term_line **last) {
+        assert(line);
+        assert(first);
+        assert(last);
+
+        if (line->lines_prev)
+                line->lines_prev->lines_next = line->lines_next;
+        else
+                *first = line->lines_next;
+        if (line->lines_next)
+                line->lines_next->lines_prev = line->lines_prev;
+        else
+                *last = line->lines_prev;
+
+        line->lines_prev = NULL;
+        line->lines_next = NULL;
+}