xref: /openbsd/usr.bin/vi/cl/README.signal (revision 404b540a)
1#	$OpenBSD: README.signal,v 1.2 2001/01/29 01:58:26 niklas Exp $
2
3#	@(#)README.signal	10.1 (Berkeley) 6/23/95
4
5There are six (normally) asynchronous actions about which vi cares:
6SIGHUP, SIGINT, SIGQUIT, SIGTERM, SIGTSTP and SIGWINCH.
7
8The assumptions:
9	1: The DB routines are not reentrant.
10	2: The curses routines may not be reentrant.
11	3: Neither DB nor curses will restart system calls.
12
13XXX
14Note, most C library functions don't restart system calls.  So, we should
15*probably* start blocking around any imported function that we don't know
16doesn't make a system call.  This is going to be a genuine annoyance...
17
18SIGHUP, SIGTERM
19	Used for file recovery.  The DB routines can't be reentered, nor
20	can they handle interrupted system calls, so the vi routines that
21	call DB block signals.  This means that DB routines could be
22	called at interrupt time, if necessary.
23
24SIGQUIT
25	Disabled by the signal initialization routines.  Historically, ^\
26	switched vi into ex mode, and we continue that practice.
27
28SIGWINCH:
29	The interrupt routine sets a global bit which is checked by the
30 	key-read routine, so there are no reentrancy issues.  This means
31	that the screen will not resize until vi runs out of keys, but
32	that doesn't seem like a problem.
33
34SIGINT and SIGTSTP are a much more difficult issue to resolve.  Vi has
35to permit the user to interrupt long-running operations.  Generally, a
36search, substitution or read/write is done on a large file, or, the user
37creates a key mapping with an infinite loop.  This problem will become
38worse as more complex semantics are added to vi, especially things like
39making it a pure text widget.  There are four major solutions on the table,
40each of which have minor permutations.
41
421:	Run in raw mode.
43
44	The up side is that there's no asynchronous behavior to worry about,
45	and obviously no reentrancy problems.  The down side is that it's easy
46	to misinterpret characters (e.g. :w big_file^Mi^V^C is going to look
47	like an interrupt) and it's easy to get into places where we won't see
48	interrupt characters (e.g. ":map a ixx^[hxxaXXX" infinitely loops in
49	historic implementations of vi).  Periodically reading the terminal
50	input buffer might solve the latter problem, but it's not going to be
51	pretty.
52
53	Also, we're going to be checking for ^C's and ^Z's both, all over
54	the place -- I hate to litter the source code with that.  For example,
55	the historic version of vi didn't permit you to suspend the screen if
56	you were on the colon command line.  This isn't right.  ^Z isn't a vi
57	command, it's a terminal event.  (Dammit.)
58
592:	Run in cbreak mode.  There are two problems in this area.  First, the
60	current curses implementations (both System V and Berkeley) don't give
61	you clean cbreak modes. For example, the IEXTEN bit is left on, turning
62	on DISCARD and LNEXT.  To clarify, what vi WANTS is 8-bit clean, with
63	the exception that flow control and signals are turned on, and curses
64	cbreak mode doesn't give you this.
65
66	We can either set raw mode and twiddle the tty, or cbreak mode and
67	twiddle the tty.  I chose to use raw mode, on the grounds that raw
68	mode is better defined and I'm less likely to be surprised by a curses
69	implementation down the road.  The twiddling consists of setting ISIG,
70	IXON/IXOFF, and disabling some of the interrupt characters (see the
71	comments in cl_init.c).  This is all found in historic System V (SVID
72	3) and POSIX 1003.1-1992, so it should be fairly portable.
73
74	The second problem is that vi permits you to enter literal signal
75	characters, e.g. ^V^C.  There are two possible solutions.  First, you
76	can turn off signals when you get a ^V, but that means that a network
77	packet containing ^V and ^C will lose, since the ^C may take effect
78	before vi reads the ^V.  (This is particularly problematic if you're
79	talking over a protocol that recognizes signals locally and sends OOB
80	packets when it sees them.)  Second, you can turn the ^C into a literal
81	character in vi, but that means that there's a race between entering
82	^V<character>^C, i.e. the sequence may end up being ^V^C<character>.
83	Also, the second solution doesn't work for flow control characters, as
84	they aren't delivered to the program as signals.
85
86	Generally, this is what historic vi did.  (It didn't have the curses
87	problems because it didn't use curses.)  It entered signals following
88	^V characters into the input stream, (which is why there's no way to
89	enter a literal flow control character).
90
913:	Run in mostly raw mode; turn signals on when doing an operation the
92	user might want to interrupt, but leave them off most of the time.
93
94	This works well for things like file reads and writes.  This doesn't
95	work well for trying to detect infinite maps.  The problem is that
96	you can write the code so that you don't have to turn on interrupts
97	per keystroke, but the code isn't pretty and it's hard to make sure
98	that an optimization doesn't cover up an infinite loop.  This also
99	requires interaction or state between the vi parser and the key
100	reading routines, as an infinite loop may still be returning keys
101	to the parser.
102
103	Also, if the user inserts an interrupt into the tty queue while the
104	interrupts are turned off, the key won't be treated as an interrupt,
105	and requiring the user to pound the keyboard to catch an interrupt
106	window is nasty.
107
1084:	Run in mostly raw mode, leaving signals on all of the time.  Done
109	by setting raw mode, and twiddling the tty's termios ISIG bit.
110
111	This works well for the interrupt cases, because the code only has
112	to check to see if the interrupt flag has been set, and can otherwise
113	ignore signals.  It's also less likely that we'll miss a case, and we
114	don't have to worry about synchronizing between the vi parser and the
115	key read routines.
116
117	The down side is that we have to turn signals off if the user wants
118	to enter a literal character (e.g. ^V^C).  If the user enters the
119	combination fast enough, or as part of a single network packet,
120	the text input routines will treat it as a signal instead of as a
121	literal character.  To some extent, we have this problem already,
122	since we turn off flow control so that the user can enter literal
123	XON/XOFF characters.
124
125	This is probably the easiest to code, and provides the smoothest
126	programming interface.
127
128There are a couple of other problems to consider.
129
130First, System V's curses doesn't handle SIGTSTP correctly.  If you use the
131newterm() interface, the TSTP signal will leave you in raw mode, and the
132final endwin() will leave you in the correct shell mode.  If you use the
133initscr() interface, the TSTP signal will return you to the correct shell
134mode, but the final endwin() will leave you in raw mode.  There you have
135it: proof that drug testing is not making any significant headway in the
136computer industry.  The 4BSD curses is deficient in that it does not have
137an interface to the terminal keypad.  So, regardless, we have to do our
138own SIGTSTP handling.
139
140The problem with this is that if we do our own SIGTSTP handling, in either
141models #3 or #4, we're going to have to call curses routines at interrupt
142time, which means that we might be reentering curses, which is something we
143don't want to do.
144
145Second, SIGTSTP has its own little problems.  It's broadcast to the entire
146process group, not sent to a single process.  The scenario goes something
147like this: the shell execs the mail program, which execs vi.  The user hits
148^Z, and all three programs get the signal, in some random order.  The mail
149program goes to sleep immediately (since it probably didn't have a SIGTSTP
150handler in place).  The shell gets a SIGCHLD, does a wait, and finds out
151that the only child in its foreground process group (of which it's aware)
152is asleep.  It then optionally resets the terminal (because the modes aren't
153how it left them), and starts prompting the user for input.  The problem is
154that somewhere in the middle of all of this, vi is resetting the terminal,
155and getting ready to send a SIGTSTP to the process group in order to put
156itself to sleep.  There's a solution to all of this: when vi starts, it puts
157itself into its own process group, and then only it (and possible child
158processes) receive the SIGTSTP.  This permits it to clean up the terminal
159and switch back to the original process group, where it sends that process
160group a SIGTSTP, putting everyone to sleep and waking the shell.
161
162Third, handing SIGTSTP asynchronously is further complicated by the child
163processes vi may fork off.  If vi calls ex, ex resets the terminal and
164starts running some filter, and SIGTSTP stops them both, vi has to know
165when it restarts that it can't repaint the screen until ex's child has
166finished running.  This is solveable, but it's annoying.
167
168Well, somebody had to make a decision, and this is the way it's going to be
169(unless I get talked out of it).  SIGINT is handled asynchronously, so
170that we can pretty much guarantee that the user can interrupt any operation
171at any time.  SIGTSTP is handled synchronously, so that we don't have to
172reenter curses and so that we don't have to play the process group games.
173^Z is recognized in the standard text input and command modes.  (^Z should
174also be recognized during operations that may potentially take a long time.
175The simplest solution is probably to twiddle the tty, install a handler for
176SIGTSTP, and then restore normal tty modes when the operation is complete.)
177