Fix "$@$@" corner case

At <https://austingroupbugs.net/view.php?id=1852>, Steffen Nurpmeso
(@sdaoden) writes:
>    twox() { one "$@$@"; }
[...]
> shells will give different results for the argument-less call to
> twox(). (Here on my box bash(1) is the only one which expands
> "$@$@" to nothing.)

Geoff Clare responds:
> The standard clearly requires "$@$@" to generate zero fields if
> there are no positional parameters. The quoted text "if the
> expansion is embedded within a word which contains one or more
> other parts that expand to a quoted null string, ..." does not
> apply because there are no other parts that expand to a quoted
> null string. (The second $@ is not a candidate for being such a
> part because "the expansion of '@' shall generate zero fields".)

Simple reproducer:

$ set --  # ensure zero PPs
$ set -- "$@$@"
$ echo $#
1

Expected output: 0

ksh has further inconsistent behaviour: e.g., for zero positional
parameters, ''"$@" correctly yields one empty field, ''"$@$@"
yields zero, and ''"$@$@$@" and on yield one again, but
''''"$@$@$@" yields zero -- etc.

Analysis: In copyto() in macro.c:705, the mp->quoted variable is
increased by one for each shell quote encountered in a word. In
varsub() in macro.c:1888, this variable is decreased by two if a
quoted $@ is encountered with zero positional parameters. For a
simple "$@" which has two quotes, this results in a decrease of two
and an mp->quoted value of 0 once both quotes have been parsed.
Effectively, both quotes have "disappeared", which allows empty
removal to take effect: correct behaviour. But if another $@ exists
within the same set of quotes, that decrease by two is done again,
causing incorrect behaviour.

src/cmd/ksh93/sh/macro.c:
- Add new 'atspecial' member to Mac_t struct.
- varsub(): When applying the quote count reduction for quoted "$@"
  with zero PPs, remember the current mp->quoted count in
  mp->atspecial, and do not apply the reduction again unless
  mp->quoted has changed from mp->atspecial. This fixes the bug.

NEWS

@@ -2,6 +2,11 @@ This documents significant changes in the dev branch of ksh 93u+m.
 For full details, see the git log at: https://github.com/ksh93/ksh
 Uppercase BUG_* IDs are shell bug IDs as used by the Modernish shell library.
 
+2024-08-19:
+
+- Fixed a corner-case bug: for an empty set of positional parameters,
+  "$@$@" would generate one empty field instead of zero fields like "$@".
+
 2024-07-31:
 
 - Fixed a bug where printf %T, after having printed the time in UTC once

src/cmd/ksh93/include/version.h

@@ -18,7 +18,7 @@
 
 #define SH_RELEASE_FORK	"93u+m"		/* only change if you develop a new ksh93 fork */
 #define SH_RELEASE_SVER	"1.1.0-alpha"	/* semantic version number: https://semver.org */
-#define SH_RELEASE_DATE	"2024-08-01"	/* must be in this format for $((.sh.version)) */
+#define SH_RELEASE_DATE	"2024-08-19"	/* must be in this format for $((.sh.version)) */
 #define SH_RELEASE_CPYR	"(c) 2020-2024 Contributors to ksh " SH_RELEASE_FORK
 
 /* Scripts sometimes field-split ${.sh.version}, so don't change amount of whitespace. */

src/cmd/ksh93/sh/macro.c

@@ -60,6 +60,7 @@ typedef struct  _mac_
 	char		*ifsp;		/* pointer to IFS value */
 	int		fields;		/* number of fields */
 	short		quoted;		/* set when word has quotes */
+	short		atspecial;	/* quoted count at which special-casing for "$@" with no PPs has been applied */
 	unsigned char	ifs;		/* first byte of IFS */
 	char		atmode;		/* when processing $@ */
 	char		quote;		/* set within double quoted contexts */
@@ -1882,9 +1883,9 @@ static int varsub(Mac_t *mp)
 		if(v || c=='/' && offset>=0)
 			stkseek(stkp,offset);
 	}
-	/* check for quoted @ */
-	if(mode=='@' && mp->quote && !v && c!='-')
-		mp->quoted-=2;
+	/* discount the quotes around $@ if there are zero PPs, so that empty "$@" generates zero fields */
+	if(mode=='@' && mp->quote && !v && c!='-' && mp->atspecial != mp->quoted)
+		mp->quoted -= 2, mp->atspecial = mp->quoted;
 retry2:
 	if(v && (!nulflg || *v ) && c!='+')
 	{

src/cmd/ksh93/tests/quoting.sh

@@ -2,7 +2,7 @@
 #                                                                      #
 #               This software is part of the ast package               #
 #          Copyright (c) 1982-2011 AT&T Intellectual Property          #
-#          Copyright (c) 2020-2022 Contributors to ksh 93u+m           #
+#          Copyright (c) 2020-2024 Contributors to ksh 93u+m           #
 #                      and is licensed under the                       #
 #                 Eclipse Public License, Version 2.0                  #
 #                                                                      #
@@ -369,4 +369,33 @@ case x in
 $x) err_exit "case \$x='$x' should not match x";;
 esac
 
+# ======
+# https://austingroupbugs.net/view.php?id=1852
+# https://mail.gnu.org/archive/html/bug-bash/2024-08/msg00132.html
+unset e E q i
+for e in 3 '"$@"' 5 '"$@$@"' 7 '"$@$@$@"' 9 '"$@$@$@$@"' 11 '"$@$@$@$@$@"' 13 '"$@$@$@$@$@$@"' \
+	5 '"$@""$@"' 7 '"$@""$@""$@"' 9 '"$@""$@""$@""$@"' 11 '"$@""$@""$@""$@""$@"' 13 '"$@""$@""$@""$@""$@""$@"'
+do	[[ $e == [0-9]* ]] && i=$e && continue
+	set --  # set zero PPs
+	eval "set -- $e"
+	(($# == 0)) || err_exit "$e does not yield zero fields for zero positional parameters (got $#)"
+	set -- one two three
+	eval "set -- $e"
+	(($# == i)) || err_exit "$e does not yield $i fields for 3 positional parameters (got $#)"
+	for q in "''" '""'
+	do	for q in "$q" "$q$q" "$q$q$q" "$q$q$q$q" "$q$q$q$q$q" "$q$q$q$q$q$q"
+		do	for E in "$q$e" "$e$q" "$q$e$q"
+			do	set --  # set zero PPs
+				eval "set -- $E"
+				(($# == 1)) || err_exit "$E does not yield one field for zero positional parameters (got $#)"
+				set -- one two three
+				eval "set -- $E"
+				(($# == i)) || err_exit "$E does not yield $i fields for 3 positional parameters (got $#)"
+			done
+		done
+	done
+	((i+=2))
+done
+
+# ======
 exit $((Errors<125?Errors:125))