Upgrade one-true-awk to 7d1848cfa6b7b3bb9a7c851339626982198a57bc am: a451bd9a44 am: 9f8a9cd5c8 am: 2b34dbd4fc

Original change: https://android-review.googlesource.com/c/platform/external/one-true-awk/+/1541457

MUST ONLY BE SUBMITTED BY AUTOMERGER

Change-Id: I104ca1ea4ef83944bbe83a668c0043c75e7dd966

13 files changed, 234 insertions, 58 deletions

diff --git a/FIXES b/FIXES
index 2aeb8d0..20b4bd8 100644
--- a/FIXES
+++ b/FIXES
@@ -25,6 +25,29 @@ THIS SOFTWARE.
 This file lists all bug fixes, changes, etc., made since the AWK book
 was sent to the printers in August, 1987.
 
+December 18, 2020:
+	Fix problems converting inf and NaN values in lib.c:is_valid_number.
+	Enhance number to string conversion to do the right thing for
+	NaN and inf values.  Things are now pretty much the same as in
+	gawk.  (Found a gawk bug while we're at it.) Added a torture
+	test for these values.  Thanks to Arnold Robbins.  Allows closing
+	of PR #101.
+
+December 15, 2020:
+	Merge PR #99, which gets the right header for strcasecmp.
+	Thanks to GitHub user michaelforney.
+
+December 8, 2020:
+	Merge PR #98: Disallow hex data. Allow only +nan, -nan,
+	+inf, -inf (case independent) to give NaN and infinity values.
+	Improve things so that string to double conversion is only
+	done once, yielding something of a speedup.  This obviate
+	PR #95. Thanks to Arnold Robbins.
+
+December 3, 2020:
+	Fix to argument parsing to avoid printing spurious newlines.
+	Thanks to Todd Miller. Merges PR #97.
+
 October 13, 2020:
 	Add casts before all the calls to malloc/calloc/realloc in order
 	to get it to compile with g++. Thanks to Arnold Robbins.
diff --git a/METADATA b/METADATA
index 0c29003..833688d 100644
--- a/METADATA
+++ b/METADATA
@@ -5,11 +5,11 @@ third_party {
     type: GIT
     value: "https://github.com/onetrueawk/awk.git"
   }
-  version: "a2a41a8e359dca5295c43dc2c2408094b96c3e0c"
+  version: "7d1848cfa6b7b3bb9a7c851339626982198a57bc"
   license_type: NOTICE
   last_upgrade_date {
-    year: 2020
-    month: 11
-    day: 30
+    year: 2021
+    month: 1
+    day: 5
   }
 }
diff --git a/README.md b/README.md
index 94075fd..b8089b3 100644
--- a/README.md
+++ b/README.md
@@ -114,10 +114,6 @@ basis.  We try to get to issues and pull requests as quickly
 as we can.  Unfortunately, however, keeping this program going
 is not at the top of our priority list.
 
-_If_ you (yes, you!) are interested in taking over active maintenance of
-`awk`, please open an issue to indicate that fact, and give us a little bit of
-your background and some idea of your plans and dreams.  Thanks!
-
 #### Last Updated
 
-Tue Oct 13 20:00:09 IDT 2020
+Fri Dec 25 16:53:34 EST 2020
diff --git a/awk.1 b/awk.1
index 050d253..d27dbff 100644
--- a/awk.1
+++ b/awk.1
@@ -577,3 +577,56 @@ The scope rules for variables in functions are a botch;
 the syntax is worse.
 .PP
 Only eight-bit characters sets are handled correctly.
+.SH UNUSUAL FLOATING-POINT VALUES
+.I Awk
+was designed before IEEE 754 arithmetic defined Not-A-Number (NaN)
+and Infinity values, which are supported by all modern floating-point
+hardware.
+.PP
+Because
+.I awk
+uses
+.IR strtod (3)
+and
+.IR atof (3)
+to convert string values to double-precision floating-point values,
+modern C libraries also convert strings starting with
+.B inf
+and
+.B nan
+into infinity and NaN values respectively.  This led to strange results,
+with something like this:
+.PP
+.EX
+.nf
+echo nancy | awk '{ print $1 + 0 }'
+.fi
+.EE
+.PP
+printing
+.B nan
+instead of zero.
+.PP
+.I Awk
+now follows GNU AWK, and prefilters string values before attempting
+to convert them to numbers, as follows:
+.TP
+.I "Hexadecimal values"
+Hexadecimal values (allowed since C99) convert to zero, as they did
+prior to C99.
+.TP
+.I "NaN values"
+The two strings
+.B +nan
+and
+.B \-nan
+(case independent) convert to NaN. No others do.
+(NaNs can have signs.)
+.TP
+.I "Infinity values"
+The two strings
+.B +inf
+and
+.B \-inf
+(case independent) convert to positive and negative infinity, respectively.
+No others do.
diff --git a/bugs-fixed/inf-nan-torture.awk b/bugs-fixed/inf-nan-torture.awk
new file mode 100644
index 0000000..8d145f2
--- /dev/null
+++ b/bugs-fixed/inf-nan-torture.awk
@@ -0,0 +1,4 @@
+{
+	for (i = 1; i <= NF; i++)
+		print i, $i, $i + 0
+}
diff --git a/bugs-fixed/inf-nan-torture.in b/bugs-fixed/inf-nan-torture.in
new file mode 100644
index 0000000..45dfdc8
--- /dev/null
+++ b/bugs-fixed/inf-nan-torture.in
@@ -0,0 +1 @@
+-inf -inform inform -nan -nancy nancy -123 0 123 +123 nancy +nancy +nan inform +inform +inf
diff --git a/bugs-fixed/inf-nan-torture.ok b/bugs-fixed/inf-nan-torture.ok
new file mode 100644
index 0000000..40d3194
--- /dev/null
+++ b/bugs-fixed/inf-nan-torture.ok
@@ -0,0 +1,16 @@
+1 -inf -inf
+2 -inform 0
+3 inform 0
+4 -nan -nan
+5 -nancy 0
+6 nancy 0
+7 -123 -123
+8 0 0
+9 123 123
+10 +123 123
+11 nancy 0
+12 +nancy 0
+13 +nan +nan
+14 inform 0
+15 +inform 0
+16 +inf +inf
diff --git a/lex.c b/lex.c
index df39c8e..9d1ae06 100644
--- a/lex.c
+++ b/lex.c
@@ -191,7 +191,12 @@ int yylex(void)
 			return word(buf);
 		if (isdigit(c)) {
 			char *cp = tostring(buf);
-			yylval.cp = setsymtab(buf, cp, atof(buf), CON|NUM, symtab);
+			double result;
+
+			if (is_number(cp, & result))
+				yylval.cp = setsymtab(buf, cp, result, CON|NUM, symtab);
+			else
+				yylval.cp = setsymtab(buf, cp, 0.0, STR, symtab);
 			free(cp);
 			/* should this also have STR set? */
 			RET(NUMBER);
diff --git a/lib.c b/lib.c
index f55e897..e8310a9 100644
--- a/lib.c
+++ b/lib.c
@@ -25,11 +25,13 @@ THIS SOFTWARE.
 #define DEBUG
 #include <stdio.h>
 #include <string.h>
+#include <strings.h>
 #include <ctype.h>
 #include <errno.h>
 #include <stdlib.h>
 #include <stdarg.h>
 #include <limits.h>
+#include <math.h>
 #include "awk.h"
 
 char	EMPTY[] = { '\0' };
@@ -181,12 +183,14 @@ int getrec(char **pbuf, int *pbufsize, bool isrecord)	/* get next input record *
 			innew = false;
 		if (c != 0 || buf[0] != '\0') {	/* normal record */
 			if (isrecord) {
+				double result;
+
 				if (freeable(fldtab[0]))
 					xfree(fldtab[0]->sval);
 				fldtab[0]->sval = buf;	/* buf == record */
 				fldtab[0]->tval = REC | STR | DONTFREE;
-				if (is_number(fldtab[0]->sval)) {
-					fldtab[0]->fval = atof(fldtab[0]->sval);
+				if (is_number(fldtab[0]->sval, & result)) {
+					fldtab[0]->fval = result;
 					fldtab[0]->tval |= NUM;
 				}
 			}
@@ -293,6 +297,7 @@ void setclvar(char *s)	/* set var=value from s */
 {
 	char *p;
 	Cell *q;
+	double result;
 
 	for (p=s; *p != '='; p++)
 		;
@@ -300,8 +305,8 @@ void setclvar(char *s)	/* set var=value from s */
 	p = qstring(p, '\0');
 	q = setsymtab(s, p, 0.0, STR, symtab);
 	setsval(q, p);
-	if (is_number(q->sval)) {
-		q->fval = atof(q->sval);
+	if (is_number(q->sval, & result)) {
+		q->fval = result;
 		q->tval |= NUM;
 	}
 	DPRINTF("command line set %s to |%s|\n", s, p);
@@ -402,9 +407,11 @@ void fldbld(void)	/* create fields from current record */
 	lastfld = i;
 	donefld = true;
 	for (j = 1; j <= lastfld; j++) {
+		double result;
+
 		p = fldtab[j];
-		if(is_number(p->sval)) {
-			p->fval = atof(p->sval);
+		if(is_number(p->sval, & result)) {
+			p->fval = result;
 			p->tval |= NUM;
 		}
 	}
@@ -669,12 +676,11 @@ void error()
 			fprintf(stderr, " source line number %d", curnode->lineno);
 		else if (lineno)
 			fprintf(stderr, " source line number %d", lineno);
+		if (compile_time == COMPILING && cursource() != NULL)
+			fprintf(stderr, " source file %s", cursource());
+		fprintf(stderr, "\n");
+		eprint();
 	}
-
-	if (compile_time == COMPILING && cursource() != NULL)
-		fprintf(stderr, " source file %s", cursource());
-	fprintf(stderr, "\n");
-	eprint();
 }
 
 void eprint(void)	/* try to print context around error */
@@ -757,24 +763,69 @@ int isclvar(const char *s)	/* is s of form var=something ? */
 /* strtod is supposed to be a proper test of what's a valid number */
 /* appears to be broken in gcc on linux: thinks 0x123 is a valid FP number */
 /* wrong: violates 4.10.1.4 of ansi C standard */
+
 /* well, not quite. As of C99, hex floating point is allowed. so this is
- * a bit of a mess.
+ * a bit of a mess. We work around the mess by checking for a hexadecimal
+ * value and disallowing it. Similarly, we now follow gawk and allow only
+ * +nan, -nan, +inf, and -inf for NaN and infinity values.
  */
 
-#include <math.h>
-int is_number(const char *s)
+/*
+ * This routine now has a more complicated interface, the main point
+ * being to avoid the double conversion of a string to double, and
+ * also to convey out, if requested, the information that the numeric
+ * value was a leading string or is all of the string. The latter bit
+ * is used in getfval().
+ */
+
+bool is_valid_number(const char *s, bool trailing_stuff_ok,
+			bool *no_trailing, double *result)
 {
 	double r;
 	char *ep;
+	bool retval = false;
+	bool is_nan = false;
+	bool is_inf = false;
+
+	if (no_trailing)
+		*no_trailing = false;
+
+	while (isspace(*s))
+		s++;
+
+	// no hex floating point, sorry
+	if (s[0] == '0' && tolower(s[1]) == 'x')
+		return false;
+
+	// allow +nan, -nan, +inf, -inf, any other letter, no
+	if (s[0] == '+' || s[0] == '-') {
+		is_nan = (strncasecmp(s+1, "nan", 3) == 0);
+		is_inf = (strncasecmp(s+1, "inf", 3) == 0);
+		if ((is_nan || is_inf)
+		    && (isspace(s[4]) || s[4] == '\0'))
+			goto convert;
+		else if (! isdigit(s[1]) && s[1] != '.')
+			return false;
+	}
+	else if (! isdigit(s[0]) && s[0] != '.')
+		return false;
+
+convert:
 	errno = 0;
 	r = strtod(s, &ep);
-	if (ep == s || r == HUGE_VAL || errno == ERANGE)
-		return 0;
-	/* allow \r as well. windows files aren't going to go away. */
-	while (*ep == ' ' || *ep == '\t' || *ep == '\n' || *ep == '\r')
-		ep++;
-	if (*ep == '\0')
-		return 1;
-	else
-		return 0;
+	if (ep == s || errno == ERANGE)
+		return false;
+
+	if (isnan(r) && s[0] == '-' && signbit(r) == 0)
+		r = -r;
+
+	if (result != NULL)
+		*result = r;
+
+	retval = (isspace(*ep) || *ep == '\0' || trailing_stuff_ok);
+
+	if (no_trailing != NULL)
+		*no_trailing = (*ep == '\0');
+
+	return retval;
 }
diff --git a/main.c b/main.c
index 6453981..5970cb4 100644
--- a/main.c
+++ b/main.c
@@ -22,7 +22,7 @@ ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF
 THIS SOFTWARE.
 ****************************************************************/
 
-const char	*version = "version 20201013";
+const char	*version = "version 20201218";
 
 #define DEBUG
 #include <stdio.h>
diff --git a/proto.h b/proto.h
index 71e3cd9..a64991b 100644
--- a/proto.h
+++ b/proto.h
@@ -146,7 +146,9 @@ extern	void	eprint(void);
 extern	void	bclass(int);
 extern	double	errcheck(double, const char *);
 extern	int	isclvar(const char *);
-extern	int	is_number(const char *);
+extern	bool	is_valid_number(const char *s, bool trailing_stuff_ok,
+				bool *no_trailing, double *result);
+#define is_number(s, val)	is_valid_number(s, false, NULL, val)
 
 extern	int	adjbuf(char **pb, int *sz, int min, int q, char **pbp, const char *what);
 extern	void	run(Node *);
diff --git a/run.c b/run.c
index 854463f..da4f555 100644
--- a/run.c
+++ b/run.c
@@ -407,6 +407,7 @@ Cell *awkgetline(Node **a, int n)	/* get next line from specific input */
 	int bufsize = recsize;
 	int mode;
 	bool newflag;
+	double result;
 
 	if ((buf = (char *) malloc(bufsize)) == NULL)
 		FATAL("out of memory in getline");
@@ -429,15 +430,15 @@ Cell *awkgetline(Node **a, int n)	/* get next line from specific input */
 		} else if (a[0] != NULL) {	/* getline var <file */
 			x = execute(a[0]);
 			setsval(x, buf);
-			if (is_number(x->sval)) {
-				x->fval = atof(x->sval);
+			if (is_number(x->sval, & result)) {
+				x->fval = result;
 				x->tval |= NUM;
 			}
 			tempfree(x);
 		} else {			/* getline <file */
 			setsval(fldtab[0], buf);
-			if (is_number(fldtab[0]->sval)) {
-				fldtab[0]->fval = atof(fldtab[0]->sval);
+			if (is_number(fldtab[0]->sval, & result)) {
+				fldtab[0]->fval = result;
 				fldtab[0]->tval |= NUM;
 			}
 		}
@@ -448,8 +449,8 @@ Cell *awkgetline(Node **a, int n)	/* get next line from specific input */
 			n = getrec(&buf, &bufsize, false);
 			x = execute(a[0]);
 			setsval(x, buf);
-			if (is_number(x->sval)) {
-				x->fval = atof(x->sval);
+			if (is_number(x->sval, & result)) {
+				x->fval = result;
 				x->tval |= NUM;
 			}
 			tempfree(x);
@@ -726,7 +727,7 @@ Cell *indirect(Node **a, int n)	/* $( a[0] ) */
 	if ((Awkfloat)INT_MAX < val)
 		FATAL("trying to access out of range field %s", x->nval);
 	m = (int) val;
-	if (m == 0 && !is_number(s = getsval(x)))	/* suspicion! */
+	if (m == 0 && !is_number(s = getsval(x), NULL))	/* suspicion! */
 		FATAL("illegal field $(%s), name \"%s\"", s, x->nval);
 		/* BUG: can x->nval ever be null??? */
 	tempfree(x);
@@ -1259,6 +1260,7 @@ Cell *split(Node **a, int nnn)	/* split(a[0], a[1], a[2]); a[3] is type */
 	int sep;
 	char temp, num[50];
 	int n, tempstat, arg3type;
+	double result;
 
 	y = execute(a[0]);	/* source string */
 	origs = s = strdup(getsval(y));
@@ -1303,8 +1305,8 @@ Cell *split(Node **a, int nnn)	/* split(a[0], a[1], a[2]); a[3] is type */
 				snprintf(num, sizeof(num), "%d", n);
 				temp = *patbeg;
 				setptr(patbeg, '\0');
-				if (is_number(s))
-					setsymtab(num, s, atof(s), STR|NUM, (Array *) ap->sval);
+				if (is_number(s, & result))
+					setsymtab(num, s, result, STR|NUM, (Array *) ap->sval);
 				else
 					setsymtab(num, s, 0.0, STR, (Array *) ap->sval);
 				setptr(patbeg, temp);
@@ -1322,8 +1324,8 @@ Cell *split(Node **a, int nnn)	/* split(a[0], a[1], a[2]); a[3] is type */
 		}
 		n++;
 		snprintf(num, sizeof(num), "%d", n);
-		if (is_number(s))
-			setsymtab(num, s, atof(s), STR|NUM, (Array *) ap->sval);
+		if (is_number(s, & result))
+			setsymtab(num, s, result, STR|NUM, (Array *) ap->sval);
 		else
 			setsymtab(num, s, 0.0, STR, (Array *) ap->sval);
   spdone:
@@ -1343,8 +1345,8 @@ Cell *split(Node **a, int nnn)	/* split(a[0], a[1], a[2]); a[3] is type */
 			temp = *s;
 			setptr(s, '\0');
 			snprintf(num, sizeof(num), "%d", n);
-			if (is_number(t))
-				setsymtab(num, t, atof(t), STR|NUM, (Array *) ap->sval);
+			if (is_number(t, & result))
+				setsymtab(num, t, result, STR|NUM, (Array *) ap->sval);
 			else
 				setsymtab(num, t, 0.0, STR, (Array *) ap->sval);
 			setptr(s, temp);
@@ -1372,8 +1374,8 @@ Cell *split(Node **a, int nnn)	/* split(a[0], a[1], a[2]); a[3] is type */
 			temp = *s;
 			setptr(s, '\0');
 			snprintf(num, sizeof(num), "%d", n);
-			if (is_number(t))
-				setsymtab(num, t, atof(t), STR|NUM, (Array *) ap->sval);
+			if (is_number(t, & result))
+				setsymtab(num, t, result, STR|NUM, (Array *) ap->sval);
 			else
 				setsymtab(num, t, 0.0, STR, (Array *) ap->sval);
 			setptr(s, temp);
diff --git a/tran.c b/tran.c
index 79964cd..add9d85 100644
--- a/tran.c
+++ b/tran.c
@@ -129,9 +129,11 @@ void arginit(int ac, char **av)	/* set up ARGV and ARGC */
 	free(cp->sval);
 	cp->sval = (char *) ARGVtab;
 	for (i = 0; i < ac; i++) {
+		double result;
+
 		sprintf(temp, "%d", i);
-		if (is_number(*av))
-			setsymtab(temp, *av, atof(*av), STR|NUM, ARGVtab);
+		if (is_number(*av, & result))
+			setsymtab(temp, *av, result, STR|NUM, ARGVtab);
 		else
 			setsymtab(temp, *av, 0.0, STR, ARGVtab);
 		av++;
@@ -148,13 +150,15 @@ void envinit(char **envp)	/* set up ENVIRON variable */
 	free(cp->sval);
 	cp->sval = (char *) ENVtab;
 	for ( ; *envp; envp++) {
+		double result;
+
 		if ((p = strchr(*envp, '=')) == NULL)
 			continue;
 		if( p == *envp ) /* no left hand side name in env string */
 			continue;
 		*p++ = 0;	/* split into two strings at = */
-		if (is_number(p))
-			setsymtab(*envp, p, atof(p), STR|NUM, ENVtab);
+		if (is_number(p, & result))
+			setsymtab(*envp, p, result, STR|NUM, ENVtab);
 		else
 			setsymtab(*envp, p, 0.0, STR, ENVtab);
 		p[-1] = '=';	/* restore in case env is passed down to a shell */
@@ -399,19 +403,36 @@ Awkfloat getfval(Cell *vp)	/* get float val of a Cell */
 	else if (isrec(vp) && !donerec)
 		recbld();
 	if (!isnum(vp)) {	/* not a number */
-		vp->fval = atof(vp->sval);	/* best guess */
-		if (is_number(vp->sval) && !(vp->tval&CON))
-			vp->tval |= NUM;	/* make NUM only sparingly */
+		double fval;
+		bool no_trailing;
+
+		if (is_valid_number(vp->sval, true, & no_trailing, & fval)) {
+			vp->fval = fval;
+			if (no_trailing && !(vp->tval&CON))
+				vp->tval |= NUM;	/* make NUM only sparingly */
+		} else
+			vp->fval = 0.0;
 	}
 	DPRINTF("getfval %p: %s = %g, t=%o\n",
 		(void*)vp, NN(vp->nval), vp->fval, vp->tval);
 	return(vp->fval);
 }
 
+static char *get_inf_nan(double d)
+{
+	if (isinf(d)) {
+		return (d < 0 ? "-inf" : "+inf");
+	} else if (isnan(d)) {
+		return (signbit(d) != 0 ? "-nan" : "+nan");
+	} else
+		return NULL;
+}
+
 static char *get_str_val(Cell *vp, char **fmt)        /* get string val of a Cell */
 {
 	char s[256];
 	double dtemp;
+	char *p;
 
 	if ((vp->tval & (NUM | STR)) == 0)
 		funnyvar(vp, "read value of");
@@ -448,7 +469,9 @@ static char *get_str_val(Cell *vp, char **fmt)        /* get string val of a Cel
 	{ \
 		if (freeable(vp)) \
 			xfree(vp->sval); \
-		if (modf(vp->fval, &dtemp) == 0)	/* it's integral */ \
+		if ((p = get_inf_nan(vp->fval)) != NULL) \
+			strcpy(s, p); \
+		else if (modf(vp->fval, &dtemp) == 0)	/* it's integral */ \
 			snprintf(s, sizeof (s), "%.30g", vp->fval); \
 		else \
 			snprintf(s, sizeof (s), *fmt, vp->fval); \