githubmirror/ncurses

doc/html/hackguide.html
962 строки39 KB
Thomas E. Dickey
ncurses 6.3 - patch 20221126
27 ноя 2022, 04:53
27 ноя 2022, 04:53205ea49
<!--
  $Id: hackguide.html,v 1.36 2022/11/26 19:31:56 tom Exp $
  ****************************************************************************
  * Copyright 2019-2020,2022 Thomas E. Dickey                                *
  * Copyright 2000-2013,2017 Free Software Foundation, Inc.                  *
  *                                                                          *
  * Permission is hereby granted, free of charge, to any person obtaining a  *
  * copy of this software and associated documentation files (the            *
  * "Software"), to deal in the Software without restriction, including      *
  * without limitation the rights to use, copy, modify, merge, publish,      *
  * distribute, distribute with modifications, sublicense, and/or sell       *
  * copies of the Software, and to permit persons to whom the Software is    *
  * furnished to do so, subject to the following conditions:                 *
  *                                                                          *
  * The above copyright notice and this permission notice shall be included  *
  * in all copies or substantial portions of the Software.                   *
  *                                                                          *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS  *
  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF               *
  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.   *
  * IN NO EVENT SHALL THE ABOVE COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,   *
  * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR    *
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR    *
  * THE USE OR OTHER DEALINGS IN THE SOFTWARE.                               *
  *                                                                          *
  * Except as contained in this notice, the name(s) of the above copyright   *
  * holders shall not be used in advertising or otherwise to promote the     *
  * sale, use or other dealings in this Software without prior written       *
  * authorization.                                                           *
  ****************************************************************************
-->
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01//EN">
<html>
<head>
  <meta name="generator" content=
  "HTML Tidy for HTML5 for Linux version 5.6.0">
  <title>A Hacker's Guide to Ncurses Internals</title>
  <link rel="author" href="mailto:bugs-ncurses@gnu.org">
  <meta http-equiv="Content-Type" content=
  "text/html; charset=us-ascii"><!--
This document is self-contained, *except* that there is one relative link to
the ncurses-intro.html document, expected to be in the same directory with
this one.
-->
</head>
<body>
  <h1 class="no-header">A Hacker's Guide to NCURSES</h1>

  <h2>A Hacker's Guide to NCURSES</h2>

  <div class="nav">
    <h2>Contents</h2>

    <ul>
      <li><a href="#abstract">Abstract</a></li>

      <li>
        <a href="#objective">Objective of the Package</a>
        <ul>
          <li><a href="#whysvr4">Why System V Curses?</a></li>

          <li><a href="#extensions">How to Design Extensions</a></li>
        </ul>
      </li>

      <li><a href="#portability">Portability and Configuration</a></li>

      <li><a href="#documentation">Documentation Conventions</a></li>

      <li><a href="#bugtrack">How to Report Bugs</a></li>

      <li>
        <a href="#ncurslib">A Tour of the Ncurses Library</a>
        <ul>
          <li><a href="#loverview">Library Overview</a></li>

          <li><a href="#engine">The Engine Room</a></li>

          <li><a href="#input">Keyboard Input</a></li>

          <li><a href="#mouse">Mouse Events</a></li>

          <li><a href="#output">Output and Screen Updating</a></li>
        </ul>
      </li>

      <li><a href="#fmnote">The Forms and Menu Libraries</a></li>

      <li>
        <a href="#tic">A Tour of the Terminfo Compiler</a>
        <ul>
          <li><a href="#nonuse">Translation of
          Non-<strong>use</strong> Capabilities</a></li>

          <li><a href="#uses">Use Capability Resolution</a></li>

          <li><a href="#translation">Source-Form Translation</a></li>
        </ul>
      </li>

      <li><a href="#utils">Other Utilities</a></li>

      <li><a href="#style">Style Tips for Developers</a></li>

      <li><a href="#port">Porting Hints</a></li>
    </ul>
  </div>

  <h2><a name="abstract" id="abstract">Abstract</a></h2>

  <p>This document is a hacker's tour of the
  <strong>ncurses</strong> library and utilities. It discusses
  design philosophy, implementation methods, and the conventions
  used for coding and documentation. It is recommended reading for
  anyone who is interested in porting, extending or improving the
  package.</p>

  <h2><a name="objective" id="objective">Objective of the
  Package</a></h2>

  <p>The objective of the <strong>ncurses</strong> package is to
  provide a free software API for character-cell terminals and
  terminal emulators with the following characteristics:</p>

  <ul>
    <li>Source-compatible with historical curses implementations
    (including the original BSD curses and System V curses.</li>

    <li>Conformant with the XSI Curses standard issued as part of
    XPG4 by X/Open.</li>

    <li>High-quality &mdash; stable and reliable code, wide
    portability, good packaging, superior documentation.</li>

    <li>Featureful &mdash; should eliminate as much of the drudgery
    of C interface programming as possible, freeing programmers to
    think at a higher level of design.</li>
  </ul>

  <p>These objectives are in priority order. So, for example,
  source compatibility with older version must trump featurefulness
  &mdash; we cannot add features if it means breaking the portion
  of the API corresponding to historical curses versions.</p>

  <h3><a name="whysvr4" id="whysvr4">Why System V Curses?</a></h3>

  <p>We used System V curses as a model, reverse-engineering their
  API, in order to fulfill the first two objectives.</p>

  <p>System V curses implementations can support BSD curses
  programs with just a recompilation, so by capturing the System V
  API we also capture BSD's.</p>

  <p>More importantly for the future, the XSI Curses standard
  issued by X/Open is explicitly and closely modeled on System V.
  So conformance with System V took us most of the way to
  base-level XSI conformance.</p>

  <h3><a name="extensions" id="extensions">How to Design
  Extensions</a></h3>

  <p>The third objective (standards conformance) requires that it
  be easy to condition source code using <strong>ncurses</strong>
  so that the absence of nonstandard extensions does not break the
  code.</p>

  <p>Accordingly, we have a policy of associating with each
  nonstandard extension a feature macro, so that ncurses client
  code can use this macro to condition in or out the code that
  requires the <strong>ncurses</strong> extension.</p>

  <p>For example, there is a macro
  <code>NCURSES_MOUSE_VERSION</code> which XSI Curses does not
  define, but which is defined in the <strong>ncurses</strong>
  library header. You can use this to condition the calls to the
  mouse API calls.</p>

  <h2><a name="portability" id="portability">Portability and
  Configuration</a></h2>

  <p>Code written for <strong>ncurses</strong> may assume an
  ANSI-standard C compiler and POSIX-compatible OS interface. It
  may also assume the presence of a System-V-compatible
  <em>select(2)</em> call.</p>

  <p>We encourage (but do not require) developers to make the code
  friendly to less-capable UNIX environments wherever possible.</p>

  <p>We encourage developers to support OS-specific optimizations
  and methods not available under POSIX/ANSI, provided only
  that:</p>

  <ul>
    <li>All such code is properly conditioned so the build process
    does not attempt to compile it under a plain ANSI/POSIX
    environment.</li>

    <li>Adding such implementation methods does not introduce
    incompatibilities in the <strong>ncurses</strong> API between
    platforms.</li>
  </ul>

  <p>We use GNU <code>autoconf(1)</code> as a tool to deal with
  portability issues. The right way to leverage an OS-specific
  feature is to modify the autoconf specification files
  (configure.in and aclocal.m4) to set up a new feature macro,
  which you then use to condition your code.</p>

  <h2><a name="documentation" id="documentation">Documentation
  Conventions</a></h2>

  <p>There are three kinds of documentation associated with this
  package. Each has a different preferred format:</p>

  <ul>
    <li>Package-internal files (README, INSTALL, TO-DO etc.)</li>

    <li>Manual pages.</li>

    <li>Everything else (i.e., narrative documentation).</li>
  </ul>

  <p>Our conventions are simple:</p>

  <ol>
    <li><strong>Maintain package-internal files in plain
    text.</strong> The expected viewer for them is <em>more(1)</em> or
    an editor window; there is no point in elaborate mark-up.</li>

    <li><strong>Mark up manual pages in the man macros.</strong>
    These have to be viewable through traditional <em>man(1)</em>
    programs.</li>

    <li><strong>Write everything else in HTML.</strong>
    </li>
  </ol>

  <p>When in doubt, HTMLize a master and use <em>lynx(1)</em> to
  generate plain ASCII (as we do for the announcement
  document).</p>

  <p>The reason for choosing HTML is that it is (a) well-adapted
  for on-line browsing through viewers that are everywhere; (b)
  more easily readable as plain text than most other mark-ups, if
  you do not have a viewer; and (c) carries enough information that
  you can generate a nice-looking printed version from it. Also, of
  course, it make exporting things like the announcement document
  to WWW pretty trivial.</p>

  <h2><a name="bugtrack" id="bugtrack">How to Report Bugs</a></h2>

  <p>The <a name="bugreport" id="bugreport">reporting address for
  bugs</a> is <a href=
  "mailto:bug-ncurses@gnu.org">bug-ncurses@gnu.org</a>. This is a
  majordomo list; to join, write to
  <code>bug-ncurses-request@gnu.org</code> with a message
  containing the line:</p>

  <pre class="code-block">
             subscribe &lt;name&gt;@&lt;host.domain&gt;
</pre>
  <p>The <code>ncurses</code> code is maintained by a small group
  of volunteers. While we try our best to fix bugs promptly, we
  simply do not have a lot of hours to spend on elementary
  hand-holding. We rely on intelligent cooperation from our users.
  If you think you have found a bug in <code>ncurses</code>, there
  are some steps you can take before contacting us that will help
  get the bug fixed quickly.</p>

  <p>In order to use our bug-fixing time efficiently, we put people
  who show us they have taken these steps at the head of our queue.
  This means that if you do not, you will probably end up at the
  tail end and have to wait a while.</p>

  <ol>
    <li><p>Develop a recipe to reproduce the bug.
      <p>Bugs we can reproduce are likely to be fixed very quickly,
      often within days. The most effective single thing you can do
      to get a quick fix is develop a way we can duplicate the bad
      behavior &mdash; ideally, by giving us source for a small,
      portable test program that breaks the library. (Even better
      is a keystroke recipe using one of the test programs provided
      with the distribution.)</p>
    </li>

    <li><p>Try to reproduce the bug on a different terminal type.
      <p>In our experience, most of the behaviors people report as
      library bugs are actually due to subtle problems in terminal
      descriptions. This is especially likely to be true if you are
      using a traditional asynchronous terminal or PC-based
      terminal emulator, rather than xterm or a UNIX console
      entry.</p>

      <p>It is therefore extremely helpful if you can tell us
      whether or not your problem reproduces on other terminal
      types. Usually you will have both a console type and xterm
      available; please tell us whether or not your bug reproduces
      on both.</p>

      <p>If you have xterm available, it is also good to collect
      xterm reports for different window sizes. This is especially
      true if you normally use an unusual xterm window size &mdash;
      a surprising number of the bugs we have seen are either
      triggered or masked by these.</p>
    </li>

    <li><p>Generate and examine a trace file for the broken behavior.
      <p>Recompile your program with the debugging versions of the
      libraries. Insert a <code>trace()</code> call with the
      argument set to <code>TRACE_UPDATE</code>. (See <a href=
      "ncurses-intro.html#debugging">"Writing Programs with
      NCURSES"</a> for details on trace levels.) Reproduce your
      bug, then look at the trace file to see what the library was
      actually doing.</p>

      <p>Another frequent cause of apparent bugs is application
      coding errors that cause the wrong things to be put on the
      virtual screen. Looking at the virtual-screen dumps in the
      trace file will tell you immediately if this is happening,
      and save you from the possible embarrassment of being told
      that the bug is in your code and is your problem rather than
      ours.</p>

      <p>If the virtual-screen dumps look correct but the bug
      persists, it is possible to crank up the trace level to give
      more and more information about the library's update actions
      and the control sequences it issues to perform them. The test
      directory of the distribution contains a tool for digesting
      these logs to make them less tedious to wade through.</p>

      <p>Often you will find terminfo problems at this stage by
      noticing that the escape sequences put out for various
      capabilities are wrong. If not, you are likely to learn
      enough to be able to characterize any bug in the
      screen-update logic quite exactly.</p>
    </li>

    <li><p>Report details and symptoms, not just interpretations.
      <p>If you do the preceding two steps, it is very likely that
      you will discover the nature of the problem yourself and be
      able to send us a fix. This will create happy feelings all
      around and earn you good karma for the first time you run
      into a bug you really cannot characterize and fix
      yourself.</p>

      <p>If you are still stuck, at least you will know what to
      tell us. Remember, we need details. If you guess about what
      is safe to leave out, you are too likely to be wrong.</p>

      <p>If your bug produces a bad update, include a trace file.
      Try to make the trace at the <em>least</em> voluminous level
      that pins down the bug. Logs that have been through
      tracemunch are OK, it does not throw away any information
      (actually they are better than un-munched ones because they
      are easier to read).</p>

      <p>If your bug produces a core-dump, please include a
      symbolic stack trace generated by gdb(1) or your local
      equivalent.</p>

      <p>Tell us about every terminal on which you have reproduced
      the bug &mdash; and every terminal on which you cannot.
      Ideally, send us terminfo sources for all of these (yours
      might differ from ours).</p>

      <p>Include your ncurses version and your OS/machine type, of
      course! You can find your ncurses version in the
      <code>curses.h</code> file.</p>
    </li>
  </ol>

  <p>If your problem smells like a logic error or in cursor
  movement or scrolling or a bad capability, there are a couple of
  tiny test frames for the library algorithms in the progs
  directory that may help you isolate it. These are not part of the
  normal build, but do have their own make productions.</p>

  <p>The most important of these is <code>mvcur</code>, a test
  frame for the cursor-movement optimization code. With this
  program, you can see directly what control sequences will be
  emitted for any given cursor movement or scroll/insert/delete
  operations. If you think you have got a bad capability
  identified, you can disable it and test again. The program is
  command-driven and has on-line help.</p>

  <p>If you think the vertical-scroll optimization is broken, or
  just want to understand how it works better, build
  <code>hashmap</code> and read the header comments of
  <code>hardscroll.c</code> and <code>hashmap.c</code>; then try it
  out. You can also test the hardware-scrolling optimization
  separately with <code>hardscroll</code>.</p>

  <h2><a name="ncurslib" id="ncurslib">A Tour of the Ncurses
  Library</a></h2>

  <h3><a name="loverview" id="loverview">Library Overview</a></h3>

  <p>Most of the library is superstructure &mdash; fairly trivial
  convenience interfaces to a small set of basic functions and data
  structures used to manipulate the virtual screen (in particular,
  none of this code does any I/O except through calls to more
  fundamental modules described below). The files</p>

  <blockquote>
    <code>lib_addch.c lib_bkgd.c lib_box.c lib_chgat.c lib_clear.c
    lib_clearok.c lib_clrbot.c lib_clreol.c lib_colorset.c
    lib_data.c lib_delch.c lib_delwin.c lib_echo.c lib_erase.c
    lib_gen.c lib_getstr.c lib_hline.c lib_immedok.c lib_inchstr.c
    lib_insch.c lib_insdel.c lib_insstr.c lib_instr.c
    lib_isendwin.c lib_keyname.c lib_leaveok.c lib_move.c
    lib_mvwin.c lib_overlay.c lib_pad.c lib_printw.c lib_redrawln.c
    lib_scanw.c lib_screen.c lib_scroll.c lib_scrollok.c
    lib_scrreg.c lib_set_term.c lib_slk.c lib_slkatr_set.c
    lib_slkatrof.c lib_slkatron.c lib_slkatrset.c lib_slkattr.c
    lib_slkclear.c lib_slkcolor.c lib_slkinit.c lib_slklab.c
    lib_slkrefr.c lib_slkset.c lib_slktouch.c lib_touch.c
    lib_unctrl.c lib_vline.c lib_wattroff.c lib_wattron.c
    lib_window.c</code>
  </blockquote>

  <p>are all in this category. They are very unlikely to need
  change, barring bugs or some fundamental reorganization in the
  underlying data structures.</p>

  <p>These files are used only for debugging support:</p>

  <blockquote>
    <code>lib_trace.c lib_traceatr.c lib_tracebits.c lib_tracechr.c
    lib_tracedmp.c lib_tracemse.c trace_buf.c</code>
  </blockquote>

  <p>It is rather unlikely you will ever need to change these,
  unless you want to introduce a new debug trace level for some
  reason.</p>

  <p>There is another group of files that do direct I/O via
  <em>tputs()</em>, computations on the terminal capabilities, or
  queries to the OS environment, but nevertheless have only fairly
  low complexity. These include:</p>

  <blockquote>
    <code>lib_acs.c lib_beep.c lib_color.c lib_endwin.c
    lib_initscr.c lib_longname.c lib_newterm.c lib_options.c
    lib_termcap.c lib_ti.c lib_tparm.c lib_tputs.c lib_vidattr.c
    read_entry.c.</code>
  </blockquote>

  <p>They are likely to need revision only if ncurses is being
  ported to an environment without an underlying terminfo
  capability representation.</p>

  <p>These files have serious hooks into the tty driver and signal
  facilities:</p>

  <blockquote>
    <code>lib_kernel.c lib_baudrate.c lib_raw.c lib_tstp.c
    lib_twait.c</code>
  </blockquote>

  <p>If you run into porting snafus moving the package to another
  UNIX, the problem is likely to be in one of these files. The file
  <code>lib_print.c</code> uses sleep(2) and also falls in this
  category.</p>

  <p>Almost all of the real work is done in the files</p>

  <blockquote>
    <code>hardscroll.c hashmap.c lib_addch.c lib_doupdate.c
    lib_getch.c lib_mouse.c lib_mvcur.c lib_refresh.c lib_setup.c
    lib_vidattr.c</code>
  </blockquote>

  <p>Most of the algorithmic complexity in the library lives in
  these files. If there is a real bug in <strong>ncurses</strong>
  itself, it is probably here. We will tour some of these files in
  detail below (see <a href="#engine">The Engine Room</a>).</p>

  <p>Finally, there is a group of files that is actually most of
  the terminfo compiler. The reason this code lives in the
  <strong>ncurses</strong> library is to support fallback to
  /etc/termcap. These files include</p>

  <blockquote>
    <code>alloc_entry.c captoinfo.c comp_captab.c comp_error.c
    comp_hash.c comp_parse.c comp_scan.c parse_entry.c
    read_termcap.c write_entry.c</code>
  </blockquote>

  <p>We will discuss these in the compiler tour.</p>

  <h3><a name="engine" id="engine">The Engine Room</a></h3>

  <h4><a name="input" id="input">Keyboard Input</a></h4>

  <p>All <code>ncurses</code> input funnels through the function
  <code>wgetch()</code>, defined in <code>lib_getch.c</code>. This
  function is tricky; it has to poll for keyboard and mouse events
  and do a running match of incoming input against the set of
  defined special keys.</p>

  <p>The central data structure in this module is a FIFO queue,
  used to match multiple-character input sequences against
  special-key capabilities; also to implement pushback via
  <code>ungetch()</code>.</p>

  <p>The <code>wgetch()</code> code distinguishes between function
  key sequences and the same sequences typed manually by doing a
  timed wait after each input character that could lead a function
  key sequence. If the entire sequence takes less than 1 second, it
  is assumed to have been generated by a function key press.</p>

  <p>Hackers bruised by previous encounters with variant
  <code>select(2)</code> calls may find the code in
  <code>lib_twait.c</code> interesting. It deals with the problem
  that some BSD selects do not return a reliable time-left value.
  The function <code>timed_wait()</code> effectively simulates a
  System V select.</p>

  <h4><a name="mouse" id="mouse">Mouse Events</a></h4>

  <p>If the mouse interface is active, <code>wgetch()</code> polls
  for mouse events each call, before it goes to the keyboard for
  input. It is up to <code>lib_mouse.c</code> how the polling is
  accomplished; it may vary for different devices.</p>

  <p>Under xterm, however, mouse event notifications come in via
  the keyboard input stream. They are recognized by having the
  <strong>kmous</strong> capability as a prefix. This is kind of
  klugey, but trying to wire in recognition of a mouse key prefix
  without going through the function-key machinery would be just
  too painful, and this turns out to imply having the prefix
  somewhere in the function-key capabilities at terminal-type
  initialization.</p>

  <p>This kluge only works because <strong>kmous</strong> is not
  actually used by any historic terminal type or curses
  implementation we know of. Best guess is it is a relic of some
  forgotten experiment in-house at Bell Labs that did not leave any
  traces in the publicly-distributed System V terminfo files. If
  System V or XPG4 ever gets serious about using it again, this
  kluge may have to change.</p>

  <p>Here are some more details about mouse event handling:</p>

  <p>The <code>lib_mouse()</code> code is logically split into a
  lower level that accepts event reports in a device-dependent
  format and an upper level that parses mouse gestures and filters
  events. The mediating data structure is a circular queue of event
  structures.</p>

  <p>Functionally, the lower level's job is to pick up primitive
  events and put them on the circular queue. This can happen in one
  of two ways: either (a) <code>_nc_mouse_event()</code> detects a
  series of incoming mouse reports and queues them, or (b) code in
  <code>lib_getch.c</code> detects the <strong>kmous</strong>
  prefix in the keyboard input stream and calls _nc_mouse_inline to
  queue up a series of adjacent mouse reports.</p>

  <p>In either case, <code>_nc_mouse_parse()</code> should be
  called after the series is accepted to parse the digested mouse
  reports (low-level events) into a gesture (a high-level or
  composite event).</p>

  <h4><a name="output" id="output">Output and Screen Updating</a></h4>

  <p>With the single exception of character echoes during a
  <code>wgetnstr()</code> call (which simulates cooked-mode line
  editing in an ncurses window), the library normally does all its
  output at refresh time.</p>

  <p>The main job is to go from the current state of the screen (as
  represented in the <code>curscr</code> window structure) to the
  desired new state (as represented in the <code>newscr</code>
  window structure), while doing as little I/O as possible.</p>

  <p>The brains of this operation are the modules
  <code>hashmap.c</code>, <code>hardscroll.c</code> and
  <code>lib_doupdate.c</code>; the latter two use
  <code>lib_mvcur.c</code>. Essentially, what happens looks like
  this:</p>

  <ul>
    <li>
      <p>The <code>hashmap.c</code> module tries to detect vertical
      motion changes between the real and virtual screens. This
      information is represented by the oldindex members in the
      newscr structure. These are modified by vertical-motion and
      clear operations, and both are re-initialized after each
      update. To this change-journalling information, the hashmap
      code adds deductions made using a modified Heckel algorithm
      on hash values generated from the line contents.</p>
    </li>

    <li>
      <p>The <code>hardscroll.c</code> module computes an optimum
      set of scroll, insertion, and deletion operations to make the
      indices match. It calls <code>_nc_mvcur_scrolln()</code> in
      <code>lib_mvcur.c</code> to do those motions.</p>
    </li>

    <li>
      <p>Then <code>lib_doupdate.c</code> goes to work. Its job is
      to do line-by-line transformations of <code>curscr</code>
      lines to <code>newscr</code> lines. Its main tool is the
      routine <code>mvcur()</code> in <code>lib_mvcur.c</code>.
      This routine does cursor-movement optimization, attempting to
      get from given screen location A to given location B in the
      fewest output characters possible.</p>
    </li>
  </ul>

  <p>If you want to work on screen optimizations, you should use
  the fact that (in the trace-enabled version of the library)
  enabling the <code>TRACE_TIMES</code> trace level causes a report
  to be emitted after each screen update giving the elapsed time
  and a count of characters emitted during the update. You can use
  this to tell when an update optimization improves efficiency.</p>

  <p>In the trace-enabled version of the library, it is also
  possible to disable and re-enable various optimizations at
  runtime by tweaking the variable
  <code>_nc_optimize_enable</code>. See the file
  <code>include/curses.h.in</code> for mask values, near the
  end.</p>

  <h2><a name="fmnote" id="fmnote">The Forms and Menu Libraries</a></h2>

  <p>The forms and menu libraries should work reliably in any
  environment you can port ncurses to. The only portability issue
  anywhere in them is what flavor of regular expressions the
  built-in form field type TYPE_REGEXP will recognize.</p>

  <p>The configuration code prefers the POSIX regex facility,
  modeled on System V's, but will settle for BSD regexps if the
  former is not available.</p>

  <p>Historical note: the panels code was written primarily to
  assist in porting u386mon 2.0 (comp.sources.misc v14i001-4) to
  systems lacking panels support; u386mon 2.10 and beyond use it.
  This version has been slightly cleaned up for
  <code>ncurses</code>.</p>

  <h2><a name="tic" id="tic">A Tour of the Terminfo Compiler</a></h2>

  <p>The <strong>ncurses</strong> implementation of
  <strong>tic</strong> is rather complex internally; it has to do a
  trying combination of missions. This starts with the fact that,
  in addition to its normal duty of compiling terminfo sources into
  loadable terminfo binaries, it has to be able to handle termcap
  syntax and compile that too into terminfo entries.</p>

  <p>The implementation therefore starts with a table-driven,
  dual-mode lexical analyzer (in <code>comp_scan.c</code>). The
  lexer chooses its mode (termcap or terminfo) based on the first
  &ldquo;,&rdquo; or &ldquo;:&rdquo; it finds in each entry. The
  lexer does all the work of recognizing capability names and
  values; the grammar above it is trivial, just "parse entries till
  you run out of file".</p>

  <h3><a name="nonuse" id="nonuse">Translation of
  Non-<strong>use</strong> Capabilities</a></h3>

  <p>Translation of most things besides <strong>use</strong>
  capabilities is pretty straightforward. The lexical analyzer's
  tokenizer hands each capability name to a hash function, which
  drives a table lookup. The table entry yields an index which is
  used to look up the token type in another table, and controls
  interpretation of the value.</p>

  <p>One possibly interesting aspect of the implementation is the
  way the compiler tables are initialized. All the tables are
  generated by various awk/sed/sh scripts from a master table
  <code>include/Caps</code>; these scripts actually write C
  initializers which are linked to the compiler. Furthermore, the
  hash table is generated in the same way, so it doesn't have to be
  generated at compiler startup time (another benefit of this
  organization is that the hash table can be in shareable text
  space).</p>

  <p>Thus, adding a new capability is usually pretty trivial, just
  a matter of adding one line to the <code>include/Caps</code>
  file. We will have more to say about this in the section on
  <a href="#translation">Source-Form Translation</a>.</p>

  <h3><a name="uses" id="uses">Use Capability Resolution</a></h3>

  <p>The background problem that makes <strong>tic</strong> tricky
  is not the capability translation itself, it is the resolution of
  <strong>use</strong> capabilities. Older versions would not
  handle forward <strong>use</strong> references for this reason
  (that is, a using terminal always had to follow its use target in
  the source file). By doing this, they got away with a simple
  implementation tactic; compile everything as it blows by, then
  resolve uses from compiled entries.</p>

  <p>This will not do for <strong>ncurses</strong>. The problem is
  that that the whole compilation process has to be embeddable in
  the <strong>ncurses</strong> library so that it can be called by
  the startup code to translate termcap entries on the fly. The
  embedded version cannot go promiscuously writing everything it
  translates out to disk &mdash; for one thing, it will typically
  be running with non-root permissions.</p>

  <p>So our <strong>tic</strong> is designed to parse an entire
  terminfo file into a doubly-linked circular list of entry
  structures in-core, and then do <strong>use</strong> resolution
  in-memory before writing everything out. This design has other
  advantages: it makes forward and back use-references equally easy
  (so we get the latter for free), and it makes checking for name
  collisions before they are written out easy to do.</p>

  <p>And this is exactly how the embedded version works. But the
  stand-alone user-accessible version of <strong>tic</strong>
  partly reverts to the historical strategy; it writes to disk (not
  keeping in core) any entry with no <strong>use</strong>
  references.</p>

  <p>This is strictly a core-economy kluge, implemented because the
  terminfo master file is large enough that some core-poor systems
  swap like crazy when you compile it all in memory...there have
  been reports of this process taking <strong>three hours</strong>,
  rather than the twenty seconds or less typical on the author's
  development box.</p>

  <p>So. The executable <strong>tic</strong> passes the
  entry-parser a hook that <em>immediately</em> writes out the
  referenced entry if it has no use capabilities. The compiler main
  loop refrains from adding the entry to the in-core list when this
  hook fires. If some other entry later needs to reference an entry
  that got written immediately, that is OK; the resolution code
  will fetch it off disk when it cannot find it in core.</p>

  <p>Name collisions will still be detected, just not as cleanly.
  The <code>write_entry()</code> code complains before overwriting
  an entry that postdates the time of <strong>tic</strong>'s first
  call to <code>write_entry()</code>, Thus it will complain about
  overwriting entries newly made during the <strong>tic</strong>
  run, but not about overwriting ones that predate it.</p>

  <h3><a name="translation" id="translation">Source-Form
  Translation</a></h3>

  <p>Another use of <strong>tic</strong> is to do source
  translation between various termcap and terminfo formats. There
  are more variants out there than you might think; the ones we
  know about are described in the <strong>captoinfo(1)</strong>
  manual page.</p>

  <p>The translation output code (<code>dump_entry()</code> in
  <code>ncurses/dump_entry.c</code>) is shared with the
  <strong>infocmp(1)</strong> utility. It takes the same internal
  representation used to generate the binary form and dumps it to
  standard output in a specified format.</p>

  <p>The <code>include/Caps</code> file has a header comment
  describing ways you can specify source translations for
  nonstandard capabilities just by altering the master table. It is
  possible to set up capability aliasing or tell the compiler to
  plain ignore a given capability without writing any C code at
  all.</p>

  <p>For circumstances where you need to do algorithmic
  translation, there are functions in <code>parse_entry.c</code>
  called after the parse of each entry that are specifically
  intended to encapsulate such translations. This, for example, is
  where the AIX <strong>box1</strong> capability get translated to
  an <strong>acsc</strong> string.</p>

  <h2><a name="utils" id="utils">Other Utilities</a></h2>

  <p>The <strong>infocmp</strong> utility is just a wrapper around
  the same entry-dumping code used by <strong>tic</strong> for
  source translation. Perhaps the one interesting aspect of the
  code is the use of a predicate function passed in to
  <code>dump_entry()</code> to control which capabilities are
  dumped. This is necessary in order to handle both the ordinary
  De-compilation case and entry difference reporting.</p>

  <p>The <strong>tput</strong> and <strong>clear</strong> utilities
  just do an entry load followed by a <code>tputs()</code> of a
  selected capability.</p>

  <h2><a name="style" id="style">Style Tips for Developers</a></h2>

  <p>See the TO-DO file in the top-level directory of the source
  distribution for additions that would be particularly useful.</p>

  <p>The prefix <code>_nc_</code> should be used on library public
  functions that are not part of the curses API in order to prevent
  pollution of the application namespace. If you have to add to or
  modify the function prototypes in curses.h.in, read
  ncurses/MKlib_gen.sh first so you can avoid breaking XSI
  conformance. Please join the ncurses mailing list. See the
  INSTALL file in the top level of the distribution for details on
  the list.</p>

  <p>Look for the string <code>FIXME</code> in source files to tag
  minor bugs and potential problems that could use fixing.</p>

  <p>Do not try to auto-detect OS features in the main body of the
  C code. That is the job of the configuration system.</p>

  <p>To hold down complexity, do make your code data-driven.
  Especially, if you can drive logic from a table filtered out of
  <code>include/Caps</code>, do it. If you find you need to augment
  the data in that file in order to generate the proper table, that
  is still preferable to ad-hoc code &mdash; that is why the fifth
  field (flags) is there.</p>

  <p>Have fun!</p>

  <h2><a name="port" id="port">Porting Hints</a></h2>

  <p>The following notes are intended to be a first step towards
  DOS and Macintosh ports of the ncurses libraries.</p>

  <p>The following library modules are &ldquo;pure curses&rdquo;;
  they operate only on the curses internal structures, do all
  output through other curses calls (not including
  <code>tputs()</code> and <code>putp()</code>) and do not call any
  other UNIX routines such as signal(2) or the stdio library. Thus,
  they should not need to be modified for single-terminal
  ports.</p>

  <blockquote>
    <code>lib_addch.c lib_addstr.c lib_bkgd.c lib_box.c lib_clear.c
    lib_clrbot.c lib_clreol.c lib_delch.c lib_delwin.c lib_erase.c
    lib_inchstr.c lib_insch.c lib_insdel.c lib_insstr.c
    lib_keyname.c lib_move.c lib_mvwin.c lib_newwin.c lib_overlay.c
    lib_pad.c lib_printw.c lib_refresh.c lib_scanw.c lib_scroll.c
    lib_scrreg.c lib_set_term.c lib_touch.c lib_tparm.c lib_tputs.c
    lib_unctrl.c lib_window.c panel.c</code>
  </blockquote>

  <p>This module is pure curses, but calls outstr():</p>

  <blockquote>
    <code>lib_getstr.c</code>
  </blockquote>

  <p>These modules are pure curses, except that they use
  <code>tputs()</code> and <code>putp()</code>:</p>

  <blockquote>
    <code>lib_beep.c lib_color.c lib_endwin.c lib_options.c
    lib_slk.c lib_vidattr.c</code>
  </blockquote>

  <p>This modules assist in POSIX emulation on non-POSIX
  systems:</p>

  <dl>
    <dt>sigaction.c</dt>

    <dd>signal calls</dd>
  </dl>

  <p>The following source files will not be needed for a
  single-terminal-type port.</p>

  <blockquote>
    <code>alloc_entry.c captoinfo.c clear.c comp_captab.c
    comp_error.c comp_hash.c comp_main.c comp_parse.c comp_scan.c
    dump_entry.c infocmp.c parse_entry.c read_entry.c tput.c
    write_entry.c</code>
  </blockquote>

  <p>The following modules will use
  open()/read()/write()/close()/lseek() on files, but no other OS
  calls.</p>

  <dl>
    <dt>lib_screen.c</dt>

    <dd>used to read/write screen dumps</dd>

    <dt>lib_trace.c</dt>

    <dd>used to write trace data to the logfile</dd>
  </dl>

  <p>Modules that would have to be modified for a port start
  here:</p>

  <p>The following modules are &ldquo;pure curses&rdquo; but
  contain assumptions inappropriate for a memory-mapped port.</p>

  <dl>
    <dt>lib_longname.c</dt>

    <dd>assumes there may be multiple terminals</dd>

    <dt>lib_acs.c</dt>

    <dd>assumes acs_map as a double indirection</dd>

    <dt>lib_mvcur.c</dt>

    <dd>assumes cursor moves have variable cost</dd>

    <dt>lib_termcap.c</dt>

    <dd>assumes there may be multiple terminals</dd>

    <dt>lib_ti.c</dt>

    <dd>assumes there may be multiple terminals</dd>
  </dl>

  <p>The following modules use UNIX-specific calls:</p>

  <dl>
    <dt>lib_doupdate.c</dt>

    <dd>input checking</dd>

    <dt>lib_getch.c</dt>

    <dd>read()</dd>

    <dt>lib_initscr.c</dt>

    <dd>getenv()</dd>

    <dt>lib_newterm.c</dt>

    <dt>lib_baudrate.c</dt>

    <dt>lib_kernel.c</dt>

    <dd>various tty-manipulation and system calls</dd>

    <dt>lib_raw.c</dt>

    <dd>various tty-manipulation calls</dd>

    <dt>lib_setup.c</dt>

    <dd>various tty-manipulation calls</dd>

    <dt>lib_restart.c</dt>

    <dd>various tty-manipulation calls</dd>

    <dt>lib_tstp.c</dt>

    <dd>signal-manipulation calls</dd>

    <dt>lib_twait.c</dt>

    <dd>gettimeofday(), select().</dd>
  </dl>

  <hr>

  <address>
    Eric S. Raymond &lt;esr@snark.thyrsus.com&gt;
  </address>
  (Note: This is <em>not</em> the <a href="#bugtrack">bug
  address</a>!)
</body>
</html>