Improve and document fast filescan loops (SHOPT_FILESCAN)

From README:

FILESCAN on  Experimental option that allows fast reading of files
             using while < file;do ...; done and allowing fields in
             each line to be accessed as positional parameters.

As SHOPT_FILESCAN has been enabled by default since ksh 93l
2001-06-01, the filescan loop is now documented in the manual page
and the compile-time option is no longer considered experimental.

We must disable this at runtime if --posix is active because it
breaks a portable use case: POSIXly, 'while <file; do stuff; done'
repeatedly excutes 'stuff' while 'file' can successfully be opened
for reading, without actually reading from 'file'.

This also backports a bugfix from the 93v- beta. Reproducer:

$ echo 'one two three' >foo
$ while <foo; do printf '[%s] ' "$@"; echo; done
[one two three]

Expected output:
[one] [two] [three]

The bug is that "$@" acts like "$*", joining all the positional
parameters into one word though it should be generating one word
for each.

src/cmd/ksh93/sh/macro.c: varsub():
- Backport fix for the bug described above. I do not understand the
  opaque macro.c code well enough yet to usefully describe the fix.

src/cmd/ksh93/sh/xec.c: sh_exec():
- Improved sanity check for filescan loop: do not recognise it if
  the simple command includes variable assignments, more than one
  redirection, or an output or append redirection.
- Disable filescan loops if --posix is active.
- Another 93v- fix: handle interrupts (errno==EINTR) when closing
  the input file.

NEWS

@@ -3,6 +3,25 @@ For full details, see the git log at: https://github.com/ksh93/ksh/tree/1.0
 
 Any uppercase BUG_* names are modernish shell bug IDs.
 
+2022-02-15:
+
+- A bug was fixed in fast filescan loops (like 'while <file; do ...; done',
+  enabled by default by the SHOPT_FILESCAN compile-time option) in which
+  "$@" incorrectly acted like "$*" within the do...done block.
+
+- Improved sanity check for filescan loops: they are now only recognised if
+  'while' is followed by a single input redirection without any command
+  name/arguments or variable assignments.
+
+- As SHOPT_FILESCAN has been enabled by default since ksh 93l 2001-06-01,
+  the fast filescan loop is now documented in the manual page and the
+  compile-time option is no longer considered experimental.
+
+- Filescan loops are now disabled at runtime if the --posix option is active
+  as they break a portable use case: POSIXly, 'while <file; do stuff; done'
+  repeatedly excutes 'stuff' while 'file' can successfully be opened for
+  reading, without actually reading from 'file'.
+
 2022-02-12:
 
 - In multibyte locales such as UTF-8, shell input is no longer corrupted when

src/cmd/ksh93/README

@@ -74,9 +74,10 @@ The options have the following defaults and meanings:
     ESH          on  Compile with emacs command line editing.  The original
                      emacs line editor code was provided by Mike Veach at IH.
 
-    FILESCAN     on  Experimental option that allows fast reading of files
-                     using while < file;do ...; done and allowing fields in
-                     each line to be accessed as positional parameters.
+    FILESCAN     on  Allows fast reading of files using:
+                         while < file; do ...; done
+                     Each line is stored in $REPLY and fields in each
+                     line can be accessed as positional parameters.
 
     FIXEDARRAY   on  When using typeset, a name in the format NAME[N]
                      creates a fixed-size array and any attempt to access a

src/cmd/ksh93/SHOPT.sh

@@ -19,7 +19,7 @@ SHOPT DYNAMIC=1				# dynamic loading for builtins
 SHOPT ECHOPRINT=			# make echo equivalent to print
 SHOPT EDPREDICT=0			# History pattern search menu (type #<pattern>, then ESC <number> TAB). Experimental.
 SHOPT ESH=1				# emacs/gmacs edit mode
-SHOPT FILESCAN=1			# fast file scan
+SHOPT FILESCAN=1			# fast file scan: while <file; do [use $REPLY or positional parameters]; done
 SHOPT FIXEDARRAY=1			# fixed dimension indexed array
 SHOPT GLOBCASEDET=			# -o globcasedetect: adapt globbing/completion to case-insensitive file systems
 SHOPT HISTEXPAND=1			# csh-style history file expansions

src/cmd/ksh93/include/shell.h

@@ -357,7 +357,6 @@ struct Shell_s
 	Shinit_f	userinit;
 	Shbltin_f	bltinfun;
 	Shbltin_t	bltindata;
-	char		*cur_line;
 	int		offsets[10];
 	Sfio_t		**sftable;
 	unsigned char	*fdstatus;
@@ -384,6 +383,9 @@ struct Shell_s
 	char 		exittrap;
 	char 		errtrap;
 	char 		end_fn;
+#if SHOPT_FILESCAN
+	char		*cur_line;
+#endif
 #if !SHOPT_DEVFD
 	char		*fifo;		/* FIFO name for current process substitution */
 	Dt_t		*fifo_tree;	/* for cleaning up process substitution FIFOs */

src/cmd/ksh93/include/version.h

@@ -21,7 +21,7 @@
 
 #define SH_RELEASE_FORK	"93u+m"		/* only change if you develop a new ksh93 fork */
 #define SH_RELEASE_SVER	"1.0.0-beta.2"	/* semantic version number: https://semver.org */
-#define SH_RELEASE_DATE	"2022-02-12"	/* must be in this format for $((.sh.version)) */
+#define SH_RELEASE_DATE	"2022-02-15"	/* must be in this format for $((.sh.version)) */
 #define SH_RELEASE_CPYR	"(c) 2020-2022 Contributors to ksh " SH_RELEASE_FORK
 
 /* Scripts sometimes field-split ${.sh.version}, so don't change amount of whitespace. */

src/cmd/ksh93/sh.1

@@ -497,6 +497,44 @@ may be used in place of
 to negate
 the loop termination test.
 .TP
+.PD 0
+\f3while\fP \f2inputredirection\^\fP \f3;do\fP \f2list\^\fP \f3;done\fP
+.PD
+Filescan loop. This is defined by a lone input redirection following
+.B while
+(see
+.I Input/Output
+below).
+It is faster than using the
+.B read
+built-in command in a regular
+.B while
+loop.
+The shell reads lines from the file or stream opened by
+.I inputredirection\^
+until the end is reached or the loop is broken.
+For each line read,
+the command
+.I list\^
+is executed with the line's contents assigned to the
+.B REPLY
+variable and the line's fields split into the positional parameters
+(see
+.I Field Splitting\^
+and
+.I Positional Parameters\^
+below).
+Within the
+.IR list\^ ,
+standard input is redirected to
+.BR /dev/null .
+If the
+.B posix
+compatibility shell option is on,
+this loop type is disabled and
+.I inputredirection\^
+is processed like a lone redirection in any other context.
+.TP
 \f3((\fP\f2expression\^\fP\f3))\fP
 .br
 The
@@ -7621,6 +7659,9 @@ causes file descriptors > 2 to be left open when invoking another program;
 .IP \[bu]
 disables the \fB&>\fR redirection shorthand;
 .IP \[bu]
+disables fast filescan loops of type
+\f3while\fP \f2inputredirection\^\fP \f3;do\fP \f2list\^\fP \f3;done\fP;
+.IP \[bu]
 makes the \fB<>\fR redirection operator default to redirecting standard input
 if no file descriptor number precedes it;
 .IP \[bu]

src/cmd/ksh93/sh/macro.c

@@ -1156,8 +1156,13 @@ retry1:
 #if  SHOPT_FILESCAN
 			if(sh.cur_line)
 			{
-				v = getdolarg(1,(int*)0);
 				dolmax = MAX_ARGN;
+				v = getdolarg(1,&vsize);
+				if(c=='*' || !mp->quoted)
+				{
+					dolmax = 1;
+					vsize = -1;
+				}
 			}
 			else
 #endif  /* SHOPT_FILESCAN */
@@ -1914,8 +1919,6 @@ retry2:
 #if  SHOPT_FILESCAN
 				if(sh.cur_line)
 				{
-					if(dolmax==MAX_ARGN && isastchar(mode))
-						break;
 					if(!(v=getdolarg(dolg,&vsize)))
 					{
 						dolmax = dolg;

src/cmd/ksh93/sh/xec.c

@@ -835,13 +835,14 @@ static void unset_instance(Namval_t *nq, Namval_t *node, struct Namref *nr,long
 #if SHOPT_FILESCAN
     static Sfio_t *openstream(struct ionod *iop, int *save)
     {
-	int savein, fd = sh_redirect(iop,3);
+	int err = errno, savein, fd = sh_redirect(iop,3);
 	Sfio_t	*sp;
 	savein = dup(0);
 	if(fd==0)
 		fd = savein;
 	sp = sfnew(NULL,NULL,SF_UNBOUND,fd,SF_READ);
-	close(0);
+	while(close(0)<0 && errno==EINTR)
+		errno = err;
 	open(e_devnull,O_RDONLY);
 	sh.offsets[0] = -1;
 	sh.offsets[1] = 0;
@@ -2242,11 +2243,17 @@ int sh_exec(register const Shnode_t *t, int flags)
 				goto endwhile;
 #endif /* SHOPT_OPTIMIZE */
 #if SHOPT_FILESCAN
-			if(type==TWH && tt->tre.tretyp==TCOM && !tt->com.comarg && tt->com.comio)
+			/* Recognize filescan loop for a lone input redirection following 'while' */
+			if(type==TWH					/* 'while' (not 'until') */
+			&& tt->tre.tretyp==TCOM 			/* one simple command follows 'while'... */
+			&& !tt->com.comarg				/* ...with no command name or arguments... */
+			&& !tt->com.comset				/* ...and no variable assignments list... */
+			&& tt->com.comio				/* ...and one I/O redirection... */
+			&& !tt->com.comio->ionxt			/* ...but not more than one... */
+			&& !(tt->com.comio->iofile & (IOPUT|IOAPP))	/* ...and not > or >> */
+			&& !sh_isoption(SH_POSIX))			/* not in POSIX compilance mode */
 			{
 				iop = openstream(tt->com.comio,&savein);
-				if(tt->com.comset)
-					nv_setlist(tt->com.comset,NV_IDENT|NV_ASSIGN,0);
 			}
 #endif /* SHOPT_FILESCAN */
 			sh.st.loopcnt++;
@@ -2291,8 +2298,10 @@ int sh_exec(register const Shnode_t *t, int flags)
 #if SHOPT_FILESCAN
 			if(iop)
 			{
+				int err=errno;
 				sfclose(iop);
-				close(0);
+				while(close(0)<0 && errno==EINTR)
+					errno = err;
 				dup(savein);
 				sh.cur_line = 0;
 			}

src/cmd/ksh93/tests/io.sh

@@ -961,5 +961,17 @@ got=$(
 [[ $got == 'test' ]] || err_exit "File descriptor is unexpectedly closed after exec in shared-state command substitution" \
 	"(expected 'test', got $(printf %q "$got"))"
 
+# ======
+# Test positional parameters in filescan loop
+# In 93u+, "$@" wrongly acted like "$*"; fix was backported from 93v- beta
+if ((SHOPT_FILESCAN))
+then
+	echo 'one/two/three' >foo
+	got=$(IFS=/; while <foo; do printf '[%s] ' "$REPLY" "$#" "$1" "$2" "$3" "$*" "$@" $* $@; done)
+	exp='[one/two/three] [3] [one] [two] [three] [one/two/three] [one] [two] [three] [one] [two] [three] [one] [two] [three] '
+	[[ $got == "$exp" ]] || err_exit '$REPLY or positional parameters incorrect in filescan loop' \
+		"(expected $(printf %q "$exp"), got $(printf %q "$got"))"
+fi
+
 # ======
 exit $((Errors<125?Errors:125))