1 A Hacker's Guide to NCURSES
6 * Objective of the Package
8 + How to Design Extensions
9 * Portability and Configuration
10 * Documentation Conventions
12 * A Tour of the Ncurses Library
17 + Output and Screen Updating
18 * The Forms and Menu Libraries
19 * A Tour of the Terminfo Compiler
20 + Translation of Non-use Capabilities
21 + Use Capability Resolution
22 + Source-Form Translation
24 * Style Tips for Developers
29 This document is a hacker's tour of the ncurses library and utilities.
30 It discusses design philosophy, implementation methods, and the
31 conventions used for coding and documentation. It is recommended
32 reading for anyone who is interested in porting, extending or
33 improving the package.
35 Objective of the Package
37 The objective of the ncurses package is to provide a free software API
38 for character-cell terminals and terminal emulators with the following
40 * Source-compatible with historical curses implementations
41 (including the original BSD curses and System V curses.
42 * Conformant with the XSI Curses standard issued as part of XPG4 by
44 * High-quality -- stable and reliable code, wide portability, good
45 packaging, superior documentation.
46 * Featureful -- should eliminate as much of the drudgery of C
47 interface programming as possible, freeing programmers to think at
48 a higher level of design.
50 These objectives are in priority order. So, for example, source
51 compatibility with older version must trump featurefulness -- we
52 cannot add features if it means breaking the portion of the API
53 corresponding to historical curses versions.
57 We used System V curses as a model, reverse-engineering their API, in
58 order to fulfill the first two objectives.
60 System V curses implementations can support BSD curses programs with
61 just a recompilation, so by capturing the System V API we also capture
64 More importantly for the future, the XSI Curses standard issued by
65 X/Open is explicitly and closely modeled on System V. So conformance
66 with System V took us most of the way to base-level XSI conformance.
68 How to Design Extensions
70 The third objective (standards conformance) requires that it be easy
71 to condition source code using ncurses so that the absence of
72 nonstandard extensions does not break the code.
74 Accordingly, we have a policy of associating with each nonstandard
75 extension a feature macro, so that ncurses client code can use this
76 macro to condition in or out the code that requires the ncurses
79 For example, there is a macro NCURSES_MOUSE_VERSION which XSI Curses
80 does not define, but which is defined in the ncurses library header.
81 You can use this to condition the calls to the mouse API calls.
83 Portability and Configuration
85 Code written for ncurses may assume an ANSI-standard C compiler and
86 POSIX-compatible OS interface. It may also assume the presence of a
87 System-V-compatible select(2) call.
89 We encourage (but do not require) developers to make the code friendly
90 to less-capable UNIX environments wherever possible.
92 We encourage developers to support OS-specific optimizations and
93 methods not available under POSIX/ANSI, provided only that:
94 * All such code is properly conditioned so the build process does
95 not attempt to compile it under a plain ANSI/POSIX environment.
96 * Adding such implementation methods does not introduce
97 incompatibilities in the ncurses API between platforms.
99 We use GNU autoconf(1) as a tool to deal with portability issues. The
100 right way to leverage an OS-specific feature is to modify the autoconf
101 specification files (configure.in and aclocal.m4) to set up a new
102 feature macro, which you then use to condition your code.
104 Documentation Conventions
106 There are three kinds of documentation associated with this package.
107 Each has a different preferred format:
108 * Package-internal files (README, INSTALL, TO-DO etc.)
110 * Everything else (i.e., narrative documentation).
112 Our conventions are simple:
113 1. Maintain package-internal files in plain text. The expected viewer
114 for them more(1) or an editor window; there's no point in
116 2. Mark up manual pages in the man macros. These have to be viewable
117 through traditional man(1) programs.
118 3. Write everything else in HTML.
120 When in doubt, HTMLize a master and use lynx(1) to generate plain
121 ASCII (as we do for the announcement document).
123 The reason for choosing HTML is that it's (a) well-adapted for on-line
124 browsing through viewers that are everywhere; (b) more easily readable
125 as plain text than most other mark-ups, if you don't have a viewer;
126 and (c) carries enough information that you can generate a
127 nice-looking printed version from it. Also, of course, it make
128 exporting things like the announcement document to WWW pretty trivial.
132 The reporting address for bugs is bug-ncurses@gnu.org. This is a
133 majordomo list; to join, write to bug-ncurses-request@gnu.org with a
134 message containing the line:
135 subscribe <name>@<host.domain>
137 The ncurses code is maintained by a small group of volunteers. While
138 we try our best to fix bugs promptly, we simply don't have a lot of
139 hours to spend on elementary hand-holding. We rely on intelligent
140 cooperation from our users. If you think you have found a bug in
141 ncurses, there are some steps you can take before contacting us that
142 will help get the bug fixed quickly.
144 In order to use our bug-fixing time efficiently, we put people who
145 show us they've taken these steps at the head of our queue. This means
146 that if you don't, you'll probably end up at the tail end and have to
148 1. Develop a recipe to reproduce the bug.
149 Bugs we can reproduce are likely to be fixed very quickly, often
150 within days. The most effective single thing you can do to get a
151 quick fix is develop a way we can duplicate the bad behavior --
152 ideally, by giving us source for a small, portable test program
153 that breaks the library. (Even better is a keystroke recipe using
154 one of the test programs provided with the distribution.)
155 2. Try to reproduce the bug on a different terminal type.
156 In our experience, most of the behaviors people report as library
157 bugs are actually due to subtle problems in terminal descriptions.
158 This is especially likely to be true if you're using a traditional
159 asynchronous terminal or PC-based terminal emulator, rather than
160 xterm or a UNIX console entry.
161 It's therefore extremely helpful if you can tell us whether or not
162 your problem reproduces on other terminal types. Usually you'll
163 have both a console type and xterm available; please tell us
164 whether or not your bug reproduces on both.
165 If you have xterm available, it is also good to collect xterm
166 reports for different window sizes. This is especially true if you
167 normally use an unusual xterm window size -- a surprising number
168 of the bugs we've seen are either triggered or masked by these.
169 3. Generate and examine a trace file for the broken behavior.
170 Recompile your program with the debugging versions of the
171 libraries. Insert a trace() call with the argument set to
172 TRACE_UPDATE. (See "Writing Programs with NCURSES" for details on
173 trace levels.) Reproduce your bug, then look at the trace file to
174 see what the library was actually doing.
175 Another frequent cause of apparent bugs is application coding
176 errors that cause the wrong things to be put on the virtual
177 screen. Looking at the virtual-screen dumps in the trace file will
178 tell you immediately if this is happening, and save you from the
179 possible embarrassment of being told that the bug is in your code
180 and is your problem rather than ours.
181 If the virtual-screen dumps look correct but the bug persists,
182 it's possible to crank up the trace level to give more and more
183 information about the library's update actions and the control
184 sequences it issues to perform them. The test directory of the
185 distribution contains a tool for digesting these logs to make them
186 less tedious to wade through.
187 Often you'll find terminfo problems at this stage by noticing that
188 the escape sequences put out for various capabilities are wrong.
189 If not, you're likely to learn enough to be able to characterize
190 any bug in the screen-update logic quite exactly.
191 4. Report details and symptoms, not just interpretations.
192 If you do the preceding two steps, it is very likely that you'll
193 discover the nature of the problem yourself and be able to send us
194 a fix. This will create happy feelings all around and earn you
195 good karma for the first time you run into a bug you really can't
196 characterize and fix yourself.
197 If you're still stuck, at least you'll know what to tell us.
198 Remember, we need details. If you guess about what is safe to
199 leave out, you are too likely to be wrong.
200 If your bug produces a bad update, include a trace file. Try to
201 make the trace at the least voluminous level that pins down the
202 bug. Logs that have been through tracemunch are OK, it doesn't
203 throw away any information (actually they're better than
204 un-munched ones because they're easier to read).
205 If your bug produces a core-dump, please include a symbolic stack
206 trace generated by gdb(1) or your local equivalent.
207 Tell us about every terminal on which you've reproduced the bug --
208 and every terminal on which you can't. Ideally, sent us terminfo
209 sources for all of these (yours might differ from ours).
210 Include your ncurses version and your OS/machine type, of course!
211 You can find your ncurses version in the curses.h file.
213 If your problem smells like a logic error or in cursor movement or
214 scrolling or a bad capability, there are a couple of tiny test frames
215 for the library algorithms in the progs directory that may help you
216 isolate it. These are not part of the normal build, but do have their
217 own make productions.
219 The most important of these is mvcur, a test frame for the
220 cursor-movement optimization code. With this program, you can see
221 directly what control sequences will be emitted for any given cursor
222 movement or scroll/insert/delete operations. If you think you've got a
223 bad capability identified, you can disable it and test again. The
224 program is command-driven and has on-line help.
226 If you think the vertical-scroll optimization is broken, or just want
227 to understand how it works better, build hashmap and read the header
228 comments of hardscroll.c and hashmap.c; then try it out. You can also
229 test the hardware-scrolling optimization separately with hardscroll.
231 A Tour of the Ncurses Library
235 Most of the library is superstructure -- fairly trivial convenience
236 interfaces to a small set of basic functions and data structures used
237 to manipulate the virtual screen (in particular, none of this code
238 does any I/O except through calls to more fundamental modules
239 described below). The files
241 lib_addch.c lib_bkgd.c lib_box.c lib_chgat.c lib_clear.c
242 lib_clearok.c lib_clrbot.c lib_clreol.c lib_colorset.c lib_data.c
243 lib_delch.c lib_delwin.c lib_echo.c lib_erase.c lib_gen.c
244 lib_getstr.c lib_hline.c lib_immedok.c lib_inchstr.c lib_insch.c
245 lib_insdel.c lib_insstr.c lib_instr.c lib_isendwin.c lib_keyname.c
246 lib_leaveok.c lib_move.c lib_mvwin.c lib_overlay.c lib_pad.c
247 lib_printw.c lib_redrawln.c lib_scanw.c lib_screen.c lib_scroll.c
248 lib_scrollok.c lib_scrreg.c lib_set_term.c lib_slk.c
249 lib_slkatr_set.c lib_slkatrof.c lib_slkatron.c lib_slkatrset.c
250 lib_slkattr.c lib_slkclear.c lib_slkcolor.c lib_slkinit.c
251 lib_slklab.c lib_slkrefr.c lib_slkset.c lib_slktouch.c lib_touch.c
252 lib_unctrl.c lib_vline.c lib_wattroff.c lib_wattron.c lib_window.c
254 are all in this category. They are very unlikely to need change,
255 barring bugs or some fundamental reorganization in the underlying data
258 These files are used only for debugging support:
260 lib_trace.c lib_traceatr.c lib_tracebits.c lib_tracechr.c
261 lib_tracedmp.c lib_tracemse.c trace_buf.c
263 It is rather unlikely you will ever need to change these, unless you
264 want to introduce a new debug trace level for some reason.
266 There is another group of files that do direct I/O via tputs(),
267 computations on the terminal capabilities, or queries to the OS
268 environment, but nevertheless have only fairly low complexity. These
271 lib_acs.c lib_beep.c lib_color.c lib_endwin.c lib_initscr.c
272 lib_longname.c lib_newterm.c lib_options.c lib_termcap.c lib_ti.c
273 lib_tparm.c lib_tputs.c lib_vidattr.c read_entry.c.
275 They are likely to need revision only if ncurses is being ported to an
276 environment without an underlying terminfo capability representation.
278 These files have serious hooks into the tty driver and signal
281 lib_kernel.c lib_baudrate.c lib_raw.c lib_tstp.c lib_twait.c
283 If you run into porting snafus moving the package to another UNIX, the
284 problem is likely to be in one of these files. The file lib_print.c
285 uses sleep(2) and also falls in this category.
287 Almost all of the real work is done in the files
289 hardscroll.c hashmap.c lib_addch.c lib_doupdate.c lib_getch.c
290 lib_mouse.c lib_mvcur.c lib_refresh.c lib_setup.c lib_vidattr.c
292 Most of the algorithmic complexity in the library lives in these
293 files. If there is a real bug in ncurses itself, it's probably here.
294 We'll tour some of these files in detail below (see The Engine Room).
296 Finally, there is a group of files that is actually most of the
297 terminfo compiler. The reason this code lives in the ncurses library
298 is to support fallback to /etc/termcap. These files include
300 alloc_entry.c captoinfo.c comp_captab.c comp_error.c comp_hash.c
301 comp_parse.c comp_scan.c parse_entry.c read_termcap.c write_entry.c
303 We'll discuss these in the compiler tour.
309 All ncurses input funnels through the function wgetch(), defined in
310 lib_getch.c. This function is tricky; it has to poll for keyboard and
311 mouse events and do a running match of incoming input against the set
312 of defined special keys.
314 The central data structure in this module is a FIFO queue, used to
315 match multiple-character input sequences against special-key
316 capabilities; also to implement pushback via ungetch().
318 The wgetch() code distinguishes between function key sequences and the
319 same sequences typed manually by doing a timed wait after each input
320 character that could lead a function key sequence. If the entire
321 sequence takes less than 1 second, it is assumed to have been
322 generated by a function key press.
324 Hackers bruised by previous encounters with variant select(2) calls
325 may find the code in lib_twait.c interesting. It deals with the
326 problem that some BSD selects don't return a reliable time-left value.
327 The function timed_wait() effectively simulates a System V select.
331 If the mouse interface is active, wgetch() polls for mouse events each
332 call, before it goes to the keyboard for input. It is up to
333 lib_mouse.c how the polling is accomplished; it may vary for different
336 Under xterm, however, mouse event notifications come in via the
337 keyboard input stream. They are recognized by having the kmous
338 capability as a prefix. This is kind of klugey, but trying to wire in
339 recognition of a mouse key prefix without going through the
340 function-key machinery would be just too painful, and this turns out
341 to imply having the prefix somewhere in the function-key capabilities
342 at terminal-type initialization.
344 This kluge only works because kmous isn't actually used by any
345 historic terminal type or curses implementation we know of. Best guess
346 is it's a relic of some forgotten experiment in-house at Bell Labs
347 that didn't leave any traces in the publicly-distributed System V
348 terminfo files. If System V or XPG4 ever gets serious about using it
349 again, this kluge may have to change.
351 Here are some more details about mouse event handling:
353 The lib_mouse()code is logically split into a lower level that accepts
354 event reports in a device-dependent format and an upper level that
355 parses mouse gestures and filters events. The mediating data structure
356 is a circular queue of event structures.
358 Functionally, the lower level's job is to pick up primitive events and
359 put them on the circular queue. This can happen in one of two ways:
360 either (a) _nc_mouse_event() detects a series of incoming mouse
361 reports and queues them, or (b) code in lib_getch.c detects the kmous
362 prefix in the keyboard input stream and calls _nc_mouse_inline to
363 queue up a series of adjacent mouse reports.
365 In either case, _nc_mouse_parse() should be called after the series is
366 accepted to parse the digested mouse reports (low-level events) into a
367 gesture (a high-level or composite event).
369 Output and Screen Updating
371 With the single exception of character echoes during a wgetnstr() call
372 (which simulates cooked-mode line editing in an ncurses window), the
373 library normally does all its output at refresh time.
375 The main job is to go from the current state of the screen (as
376 represented in the curscr window structure) to the desired new state
377 (as represented in the newscr window structure), while doing as little
380 The brains of this operation are the modules hashmap.c, hardscroll.c
381 and lib_doupdate.c; the latter two use lib_mvcur.c. Essentially, what
382 happens looks like this:
384 The hashmap.c module tries to detect vertical motion changes between
385 the real and virtual screens. This information is represented by the
386 oldindex members in the newscr structure. These are modified by
387 vertical-motion and clear operations, and both are re-initialized
388 after each update. To this change-journalling information, the hashmap
389 code adds deductions made using a modified Heckel algorithm on hash
390 values generated from the line contents.
392 The hardscroll.c module computes an optimum set of scroll, insertion,
393 and deletion operations to make the indices match. It calls
394 _nc_mvcur_scrolln() in lib_mvcur.c to do those motions.
396 Then lib_doupdate.c goes to work. Its job is to do line-by-line
397 transformations of curscr lines to newscr lines. Its main tool is the
398 routine mvcur() in lib_mvcur.c. This routine does cursor-movement
399 optimization, attempting to get from given screen location A to given
400 location B in the fewest output characters possible.
402 If you want to work on screen optimizations, you should use the fact
403 that (in the trace-enabled version of the library) enabling the
404 TRACE_TIMES trace level causes a report to be emitted after each
405 screen update giving the elapsed time and a count of characters
406 emitted during the update. You can use this to tell when an update
407 optimization improves efficiency.
409 In the trace-enabled version of the library, it is also possible to
410 disable and re-enable various optimizations at runtime by tweaking the
411 variable _nc_optimize_enable. See the file include/curses.h.in for
412 mask values, near the end.
414 The Forms and Menu Libraries
416 The forms and menu libraries should work reliably in any environment
417 you can port ncurses to. The only portability issue anywhere in them
418 is what flavor of regular expressions the built-in form field type
419 TYPE_REGEXP will recognize.
421 The configuration code prefers the POSIX regex facility, modeled on
422 System V's, but will settle for BSD regexps if the former isn't
425 Historical note: the panels code was written primarily to assist in
426 porting u386mon 2.0 (comp.sources.misc v14i001-4) to systems lacking
427 panels support; u386mon 2.10 and beyond use it. This version has been
428 slightly cleaned up for ncurses.
430 A Tour of the Terminfo Compiler
432 The ncurses implementation of tic is rather complex internally; it has
433 to do a trying combination of missions. This starts with the fact
434 that, in addition to its normal duty of compiling terminfo sources
435 into loadable terminfo binaries, it has to be able to handle termcap
436 syntax and compile that too into terminfo entries.
438 The implementation therefore starts with a table-driven, dual-mode
439 lexical analyzer (in comp_scan.c). The lexer chooses its mode (termcap
440 or terminfo) based on the first `,' or `:' it finds in each entry. The
441 lexer does all the work of recognizing capability names and values;
442 the grammar above it is trivial, just "parse entries till you run out
445 Translation of Non-use Capabilities
447 Translation of most things besides use capabilities is pretty
448 straightforward. The lexical analyzer's tokenizer hands each
449 capability name to a hash function, which drives a table lookup. The
450 table entry yields an index which is used to look up the token type in
451 another table, and controls interpretation of the value.
453 One possibly interesting aspect of the implementation is the way the
454 compiler tables are initialized. All the tables are generated by
455 various awk/sed/sh scripts from a master table include/Caps; these
456 scripts actually write C initializers which are linked to the
457 compiler. Furthermore, the hash table is generated in the same way, so
458 it doesn't have to be generated at compiler startup time (another
459 benefit of this organization is that the hash table can be in
460 shareable text space).
462 Thus, adding a new capability is usually pretty trivial, just a matter
463 of adding one line to the include/Caps file. We'll have more to say
464 about this in the section on Source-Form Translation.
466 Use Capability Resolution
468 The background problem that makes tic tricky isn't the capability
469 translation itself, it's the resolution of use capabilities. Older
470 versions would not handle forward use references for this reason (that
471 is, a using terminal always had to follow its use target in the source
472 file). By doing this, they got away with a simple implementation
473 tactic; compile everything as it blows by, then resolve uses from
476 This won't do for ncurses. The problem is that that the whole
477 compilation process has to be embeddable in the ncurses library so
478 that it can be called by the startup code to translate termcap entries
479 on the fly. The embedded version can't go promiscuously writing
480 everything it translates out to disk -- for one thing, it will
481 typically be running with non-root permissions.
483 So our tic is designed to parse an entire terminfo file into a
484 doubly-linked circular list of entry structures in-core, and then do
485 use resolution in-memory before writing everything out. This design
486 has other advantages: it makes forward and back use-references equally
487 easy (so we get the latter for free), and it makes checking for name
488 collisions before they're written out easy to do.
490 And this is exactly how the embedded version works. But the
491 stand-alone user-accessible version of tic partly reverts to the
492 historical strategy; it writes to disk (not keeping in core) any entry
493 with no use references.
495 This is strictly a core-economy kluge, implemented because the
496 terminfo master file is large enough that some core-poor systems swap
497 like crazy when you compile it all in memory...there have been reports
498 of this process taking three hours, rather than the twenty seconds or
499 less typical on the author's development box.
501 So. The executable tic passes the entry-parser a hook that immediately
502 writes out the referenced entry if it has no use capabilities. The
503 compiler main loop refrains from adding the entry to the in-core list
504 when this hook fires. If some other entry later needs to reference an
505 entry that got written immediately, that's OK; the resolution code
506 will fetch it off disk when it can't find it in core.
508 Name collisions will still be detected, just not as cleanly. The
509 write_entry() code complains before overwriting an entry that
510 postdates the time of tic's first call to write_entry(), Thus it will
511 complain about overwriting entries newly made during the tic run, but
512 not about overwriting ones that predate it.
514 Source-Form Translation
516 Another use of tic is to do source translation between various termcap
517 and terminfo formats. There are more variants out there than you might
518 think; the ones we know about are described in the captoinfo(1) manual
521 The translation output code (dump_entry() in ncurses/dump_entry.c) is
522 shared with the infocmp(1) utility. It takes the same internal
523 representation used to generate the binary form and dumps it to
524 standard output in a specified format.
526 The include/Caps file has a header comment describing ways you can
527 specify source translations for nonstandard capabilities just by
528 altering the master table. It's possible to set up capability aliasing
529 or tell the compiler to plain ignore a given capability without
530 writing any C code at all.
532 For circumstances where you need to do algorithmic translation, there
533 are functions in parse_entry.c called after the parse of each entry
534 that are specifically intended to encapsulate such translations. This,
535 for example, is where the AIX box1 capability get translated to an
540 The infocmp utility is just a wrapper around the same entry-dumping
541 code used by tic for source translation. Perhaps the one interesting
542 aspect of the code is the use of a predicate function passed in to
543 dump_entry() to control which capabilities are dumped. This is
544 necessary in order to handle both the ordinary De-compilation case and
545 entry difference reporting.
547 The tput and clear utilities just do an entry load followed by a
548 tputs() of a selected capability.
550 Style Tips for Developers
552 See the TO-DO file in the top-level directory of the source
553 distribution for additions that would be particularly useful.
555 The prefix _nc_ should be used on library public functions that are
556 not part of the curses API in order to prevent pollution of the
557 application namespace. If you have to add to or modify the function
558 prototypes in curses.h.in, read ncurses/MKlib_gen.sh first so you can
559 avoid breaking XSI conformance. Please join the ncurses mailing list.
560 See the INSTALL file in the top level of the distribution for details
563 Look for the string FIXME in source files to tag minor bugs and
564 potential problems that could use fixing.
566 Don't try to auto-detect OS features in the main body of the C code.
567 That's the job of the configuration system.
569 To hold down complexity, do make your code data-driven. Especially, if
570 you can drive logic from a table filtered out of include/Caps, do it.
571 If you find you need to augment the data in that file in order to
572 generate the proper table, that's still preferable to ad-hoc code --
573 that's why the fifth field (flags) is there.
579 The following notes are intended to be a first step towards DOS and
580 Macintosh ports of the ncurses libraries.
582 The following library modules are `pure curses'; they operate only on
583 the curses internal structures, do all output through other curses
584 calls (not including tputs() and putp()) and do not call any other
585 UNIX routines such as signal(2) or the stdio library. Thus, they
586 should not need to be modified for single-terminal ports.
588 lib_addch.c lib_addstr.c lib_bkgd.c lib_box.c lib_clear.c
589 lib_clrbot.c lib_clreol.c lib_delch.c lib_delwin.c lib_erase.c
590 lib_inchstr.c lib_insch.c lib_insdel.c lib_insstr.c lib_keyname.c
591 lib_move.c lib_mvwin.c lib_newwin.c lib_overlay.c lib_pad.c
592 lib_printw.c lib_refresh.c lib_scanw.c lib_scroll.c lib_scrreg.c
593 lib_set_term.c lib_touch.c lib_tparm.c lib_tputs.c lib_unctrl.c
596 This module is pure curses, but calls outstr():
600 These modules are pure curses, except that they use tputs() and
603 lib_beep.c lib_color.c lib_endwin.c lib_options.c lib_slk.c
606 This modules assist in POSIX emulation on non-POSIX systems:
611 The following source files will not be needed for a
612 single-terminal-type port.
614 alloc_entry.c captoinfo.c clear.c comp_captab.c comp_error.c
615 comp_hash.c comp_main.c comp_parse.c comp_scan.c dump_entry.c
616 infocmp.c parse_entry.c read_entry.c tput.c write_entry.c
618 The following modules will use open()/read()/write()/close()/lseek()
619 on files, but no other OS calls.
622 used to read/write screen dumps
625 used to write trace data to the logfile
627 Modules that would have to be modified for a port start here:
629 The following modules are `pure curses' but contain assumptions
630 inappropriate for a memory-mapped port.
633 assumes there may be multiple terminals
636 assumes acs_map as a double indirection
639 assumes cursor moves have variable cost
642 assumes there may be multiple terminals
645 assumes there may be multiple terminals
647 The following modules use UNIX-specific calls:
661 various tty-manipulation and system calls
664 various tty-manipulation calls
667 various tty-manipulation calls
670 various tty-manipulation calls
673 signal-manipulation calls
676 gettimeofday(), select().
677 _________________________________________________________________
680 Eric S. Raymond <esr@snark.thyrsus.com>
682 (Note: This is not the bug address!)