← Commit list

9958bbd70

Denys Vlasenko

2026-08-30

SUCCESS Duration: 150s

Subject

awk: disallow OFMT/CONVFMT to have more than one %SPEC, implement %s

Binary

No binary (build failed)

Architecture & Compiler Flags

Architecture: x86_64-linux-gnu Compiler: gcc (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0

gcc -Wp,-MD,applets/.usage.d -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer      -Iinclude -Iinclude -o applets/usage applets/usage.c

Build Log

(log yok)

Commit Diff GitHub →

editors/awk.c +49 -5 testsuite/awk.tests +26 -1

editors/awk.c+49 -5

@@ -967,20 +967,64 @@ static double my_strtod_or_hexoct(char **pp)

	/* -------- working with variables (set/get/copy/etc) -------- */

+	/* Perform OFMT or CONVFMT float-to-string conversion.
+	* According to POSIX, awk's behavior is undefined if OFMT
+	* contains anything but a floating-point conversion specification.
+	* However, GNU awk 5.3.0 allows prefix and suffix: "N:%.1f!"
+	* integer formats (%d et al), %s formats.
+	* It even allows %c, but we don't implement this.
+	* GNU awk also supports %ld, %lx, we don't (yet). GNU awk doesn't support %lld.
+	*/
	static const char fmt_num(const char format, double n)
	{
-	if (n == (long long)n) {
-	snprintf(g_buf, MAXVARFMT, "%lld", (long long)n);
+	if (n == trunc(n)) {
+	/* No fractionals? Try: awk -v OFMT='R:%d!' 'BEGIN { print 2*128 }' /
+	snprintf(g_buf, MAXVARFMT, "%.0f", n);
	} else {
	const char *s = format;
	char c;

-	do { c = s; } while (c && ++s);
+	/* Find %SPEC in format string */
+	for (;;) {
+	s = strchr(s, '%');
+	if (!s)
+	goto print_verbatim;
+	if (s[1] != '%')
+	break;
+	s += 2; /* skip "%%" */
+	}
+	if (strchr(s + 1, '%') != NULL)
+	//TODO: allow "%%"s in the suffix too?
+	goto INV_FMT; /* more than one %SPEC */
+
+	//TODO: factor out awk_printf machinery to make %d print large numbers too. Example:
+	// awk -v OFMT='R:%d!' 'BEGIN { print 184467440737095.1 }'
+	// R:184467440737095!
+	/* Find the format letter */
+	for (;;) {
+	c = *++s;
+	if (!c) /* unterminated "....%SPEC" */
+	goto print_verbatim;
+	if (c == '*')
+	goto INV_FMT; /* CONVFMT/OFMT='%f' is not allowed /
+	if (isalpha(c))
+	break;
+	}
+
	if (strchr("diouxX", c)) {
	snprintf(g_buf, MAXVARFMT, format, (int)n);
	} else if (strchr("eEfFgGaA", c)) {
+	print_verbatim:
	snprintf(g_buf, MAXVARFMT, format, n);
+	} else if (c == 's') {
+	/* GNU awk OFMT="%..s" formats number through "%.6g",
+	* then prints string through %s:
+	*/
+	char nbuf[sizeof(n)3 + 12]; / +5 should do? +12 is paranoia */
+	sprintf(nbuf, "%.6g", n);
+	snprintf(g_buf, MAXVARFMT, format, nbuf);
	} else {
+	INV_FMT:
	syntax_error(EMSG_INV_FMT);
	}
	}
@@ -2507,6 +2551,7 @@ static char awk_printf(node n, size_t *len)
	}
	goto tail; /* append remaining string, exit loop */
	}
+	//TODO: GNU awk 5.3.0 also supports %ld, %lx, we don't (yet). It does not support %lld.
	if (isalpha(c))
	break;
	starred \|= (c == '*');
@@ -3307,8 +3352,7 @@ static var evaluate(node op, var *res)
	for (;;) {
	var *v = evaluate(nextarg(&op1), TMPVAR0);
	if (v->type & VF_NUMBER) {
-	fputs(fmt_num(getvar_s(intvar[OFMT]), getvar_i(v)),
-	F);
+	fputs(fmt_num(getvar_s(intvar[OFMT]), getvar_i(v)), F);
	} else {
	fputs(getvar_s(v), F);
	}

testsuite/awk.tests+26 -1

@@ -75,7 +75,7 @@ testing "awk %.3d 99999999999999000.1234" \
	testing "awk %.*d 3 99999999999999000.1234" \
	"awk 'BEGIN { printf \"%.*d\n\", 3, 99999999999999000.1234; }'" "99999999999999008\n" "" ""

-	# fractioals are rounded towards 0
+	# fractionals are rounded towards 0
	testing "awk %d -1.9" \
	"awk 'BEGIN { printf \"%d\n\", -1.9; }'" "-1\n" "" ""
	testing "awk %u -1.9" \
@@ -85,6 +85,31 @@ testing "awk %x -1.9" \
	testing "awk %o -1.9" \
	"awk 'BEGIN { printf \"%o\n\", -1.9; }'" "1777777777777777777777\n" "" ""

+	# OFMT can contain literals before and after %SPEC:
+	testing "awk OFMT='R%%:%f!' -1.9" \
+	"awk -v OFMT='R%%:%f!' 'BEGIN { print -1.9; }'" "R%:-1.900000!\n" "" ""
+
+	# OFMT is ignored if the value has no fractionals (even if it does not fit into uint64)
+	testing "awk OFMT='R%%:%f!' 2**128" \
+	"awk -v OFMT='R%%:%f!' 'BEGIN { print 2**128; }'" "340282366920938463463374607431768211456\n" "" ""
+	testing "awk OFMT='R%%:%f!' 2**128" \
+	"awk -v OFMT='R%%:%f!' 'BEGIN { print -2**129; }'" "-680564733841876926926749214863536422912\n" "" ""
+
+	# OFMT can contain no %SPEC (be a literal):
+	testing "awk OFMT='No formats, %% allowed!'" \
+	"awk -v OFMT='No formats, %% allowed!' 'BEGIN { print -1.9; }'" "No formats, % allowed!\n" "" ""
+
+	# OFMT can't have more than one %SPEC:
+	testing "awk OFMT='%f%s' is not allowed" \
+	"awk -v OFMT='R%%:%f!%s!' 'BEGIN { print -1.9; }' 2>&1" \
+	"awk: cmd. line:1: Invalid format specifier\n" "" ""
+
+	# OFMT %s:
+	testing "awk OFMT='%-22s' -1.9" \
+	"awk -v OFMT='R:%-22s!' 'BEGIN { print -1.9; }'" \
+	"R:-1.9 !\n" "" ""
+
+
	# 4294967295 = 0xffffffff
	testing "awk bitwise op" "awk '{ print or(4294967295,1) }'" "4294967295\n" "" "\n"