detect struct stat.st_dev's size and signedness, and return it safely

On FreeBSD dev_t (and hence the st_dev member of struct stat) is an
unsigned 64-bit integer, and the previous simple PUSHi() corrupted
that.

A previous version of this reflected the st_ino code and implemented
our own number to string conversion, but a system with such a large
st_dev should be assumed to have inttypes.h, and an intmax_t which is
no smaller than st_dev.

The st_ino code could probably be changed similarly, but 64-bit inode
numbers are not a new thing, so it may be riskier.

Author: tonycoz

Configure

@@ -1320,6 +1320,8 @@ shsharp=''
 spitshell=''
 src=''
 ssizetype=''
+st_dev_sign=''
+st_dev_size=''
 st_ino_sign=''
 st_ino_size=''
 startperl=''
@@ -22695,6 +22697,74 @@ else
 fi
 $rm_try
 
+: Check the size of st_dev
+$echo " "
+$echo "Checking the size of st_dev..." >&4
+$cat > try.c <<EOCP
+#include <sys/stat.h>
+#include <stdio.h>
+#$i_stdlib I_STDLIB
+#ifdef I_STDLIB
+#include <stdlib.h>
+#endif
+int main() {
+    struct stat st;
+    printf("%d\n", (int)sizeof(st.st_dev));
+    exit(0);
+}
+EOCP
+set try
+if eval $compile_ok; then
+	val=`$run ./try`
+	case "$val" in
+	'')	st_dev_size=4
+		$echo "(I can't execute the test program--guessing $st_dev_size.)" >&4
+		;;
+	*)	st_dev_size=$val
+		$echo "Your st_dev is $st_dev_size bytes long."
+		;;
+	esac
+else
+	st_dev_size=4
+	$echo "(I can't compile the test program--guessing $st_dev_size.)" >&4
+fi
+$rm_try
+
+: Check if st_dev is signed
+$echo " "
+$echo "Checking the sign of st_dev..." >&4
+$cat > try.c <<EOCP
+#include <sys/stat.h>
+#include <stdio.h>
+int main() {
+	struct stat foo;
+        foo.st_dev = -1;
+	if (foo.st_dev < 0)
+		printf("-1\n");
+	else
+		printf("1\n");
+}
+EOCP
+set try
+if eval $compile; then
+	val=`$run ./try`
+	case "$val" in
+	'')	st_dev_sign=1
+		$echo "(I can't execute the test program--guessing unsigned.)" >&4
+		;;
+	*)	st_dev_sign=$val
+		case "$st_dev_sign" in
+		 1) $echo "Your st_dev is unsigned." ;;
+		-1) $echo "Your st_dev is signed."   ;;
+		esac
+		;;
+	esac
+else
+	st_dev_sign=1
+	$echo "(I can't compile the test program--guessing unsigned.)" >&4
+fi
+$rm_try
+
 : see what type of char stdio uses.
 echo " "
 echo '#include <stdio.h>' | $cppstdin $cppminus > stdioh
@@ -25250,6 +25320,8 @@ srand48_r_proto='$srand48_r_proto'
 srandom_r_proto='$srandom_r_proto'
 src='$src'
 ssizetype='$ssizetype'
+st_dev_sign='$st_dev_sign'
+st_dev_size='$st_dev_size'
 st_ino_sign='$st_ino_sign'
 st_ino_size='$st_ino_size'
 startperl='$startperl'

Cross/config.sh-arm-linux

@@ -1078,6 +1078,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src='.'
 ssizetype='ssize_t'
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='4'
 startperl='#!/usr/bin/perl'

Cross/config.sh-arm-linux-n770

@@ -1076,6 +1076,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src='.'
 ssizetype='ssize_t'
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='4'
 startperl='#!/usr/bin/perl'

NetWare/config.wc

@@ -1042,6 +1042,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src=''
 ssizetype='int'
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='4'
 startperl='#!perl'

Porting/config.sh

@@ -1107,6 +1107,8 @@ srand48_r_proto='REENTRANT_PROTO_I_LS'
 srandom_r_proto='REENTRANT_PROTO_I_TS'
 src='.'
 ssizetype='ssize_t'
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='8'
 startperl='#!/opt/perl/bin/perl5.35.2'

config_h.SH

@@ -4193,13 +4193,22 @@ sed <<!GROK!THIS! >$CONFIG_H -e 's!^#undef\(.*/\)\*!/\*#define\1 \*!' -e 's!^#un
  */
 #define SELECT_MIN_BITS	$selectminbits	/**/
 
+/* ST_DEV_SIZE:
+ *	This variable contains the size of struct stat's st_dev in bytes.
+ */
+/* ST_DEV_SIGN:
+ *	This symbol holds the signedness of struct stat's st_dev.
+ *	1 for unsigned, -1 for signed.
+ */
 /* ST_INO_SIZE:
  *	This variable contains the size of struct stat's st_ino in bytes.
  */
 /* ST_INO_SIGN:
  *	This symbol holds the signedness of struct stat's st_ino.
  *	1 for unsigned, -1 for signed.
  */
+#define ST_DEV_SIGN $st_dev_sign	/* st_dev sign */
+#define ST_DEV_SIZE $st_dev_size	/* st_dev size */
 #define ST_INO_SIGN $st_ino_sign	/* st_ino sign */
 #define ST_INO_SIZE $st_ino_size	/* st_ino size */
 

configure.com

@@ -6970,6 +6970,8 @@ $ WC "src='" + src + "'"
 $ WC "ssizetype='int'"
 $ WC "startperl=" + startperl ! This one's special--no enclosing single quotes
 $ WC "static_ext='" + static_ext + "'"
+$ WC "st_dev_size='"4"'"
+$ WC "st_dev_sign='1'"
 $ WC "st_ino_size='" + st_ino_size + "'"
 $ WC "st_ino_sign='1'"
 $ WC "stdchar='" + stdchar + "'"

plan9/config_sh.sample

@@ -1049,6 +1049,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src='.'
 ssizetype='ssize_t'
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='4'
 startperl='#!/bin/perl'

pp_sys.c

@@ -2915,7 +2915,38 @@ PP(pp_stat)
     if (max) {
         EXTEND(SP, max);
         EXTEND_MORTAL(max);
+#if ST_DEV_SIZE < IVSIZE || (ST_DEV_SIZE == IVSIZE && ST_DEV_SIGN < 0)
         mPUSHi(PL_statcache.st_dev);
+#elif ST_DEV_SIZE == IVSIZE
+        mPUSHu(PL_statcache.st_dev);
+#else
+#  if ST_DEV_SIGN < 0
+        if (LIKELY((IV)PL_statcache.st_dev == PL_statcache.st_dev)) {
+            mPUSHi((IV)PL_statcache.st_dev);
+        }
+#  else
+        if (LIKELY((UV)PL_statcache.st_dev == PL_statcache.st_dev)) {
+            mPUSHu((UV)PL_statcache.st_dev);
+        }
+#  endif
+        else {
+            char buf[sizeof(PL_statcache.st_dev)*3+1];
+            /* sv_catpvf() casts 'j' size values down to IV, so it
+               isn't suitable for use here.
+            */
+#    if defined(I_INTTYPES) && defined(HAS_SNPRINTF)
+#      if ST_DEV_SIGN < 0
+            int size = snprintf(buf, sizeof(buf), "%" PRIdMAX, (intmax_t)PL_statcache.st_dev);
+#      else
+            int size = snprintf(buf, sizeof(buf), "%" PRIuMAX, (uintmax_t)PL_statcache.st_dev);
+#      endif
+            STATIC_ASSERT_STMT(sizeof(intmax_t) >= sizeof(PL_statcache.st_dev));
+            mPUSHp(buf, size);
+#    else
+#      error extraordinarily large st_dev but no inttypes.h or no snprintf
+#    endif
+        }
+#endif
         {
             /*
              * We try to represent st_ino as a native IV or UV where

t/op/stat.t

@@ -29,7 +29,7 @@ if ($^O eq 'MSWin32') {
 
 my $Errno_loaded = eval { require Errno };
 
-plan tests => 110;
+plan tests => 111;
 
 my $Perl = which_perl();
 
@@ -664,6 +664,24 @@ SKIP: {
     unlink $link;
 }
 
+ SKIP:
+{
+    # test needs a FreeBSD /usr/bin/stat
+    # /tmp is typically tmpfs on a new FreeBSD
+    $^O eq "freebsd"
+        or skip "only checking freebsd for now", 1;
+    -x "/usr/bin/stat"
+        or skip "no /usr/bin/stat", 1;
+    my @s = stat "/tmp";
+    @s or skip "No /tmp found", 1;
+    my $test = `/usr/bin/stat -f '%d %i' /tmp`;
+    $test && $test =~ /^-?\d+ -?\d+/
+        or skip "stat didn't return an expected result";
+    chomp $test;
+    is("$s[0] $s[1]", $test,
+       "perl stat didn't match system stat utility");
+}
+
 END {
     chmod 0666, $tmpfile;
     unlink_all $tmpfile;

uconfig.h

@@ -4158,13 +4158,22 @@
  */
 #define SELECT_MIN_BITS	32	/**/
 
+/* ST_DEV_SIZE:
+ *	This variable contains the size of struct stat's st_dev in bytes.
+ */
+/* ST_DEV_SIGN:
+ *	This symbol holds the signedness of struct stat's st_dev.
+ *	1 for unsigned, -1 for signed.
+ */
 /* ST_INO_SIZE:
  *	This variable contains the size of struct stat's st_ino in bytes.
  */
 /* ST_INO_SIGN:
  *	This symbol holds the signedness of struct stat's st_ino.
  *	1 for unsigned, -1 for signed.
  */
+#define ST_DEV_SIGN 1	/* st_dev sign */
+#define ST_DEV_SIZE 4	/* st_dev size */
 #define ST_INO_SIGN 1	/* st_ino sign */
 #define ST_INO_SIZE 4	/* st_ino size */
 
@@ -5313,6 +5322,6 @@
 #endif
 
 /* Generated from:
- * 55a531381747550c11c2c61b9a9da2dacde4df465b874df55a9c923e495deb3a config_h.SH
- * 2fece1e405c60ae089fe55acaa42471b6fba78b7ab4cefc6d5e18a94b72fc2c4 uconfig.sh
+ * e953597150b0e47d177abc2f4dd1351a49557e0b537ff593acf85f76709bcce7 config_h.SH
+ * 801886c55550c022c8edfea3ef6873294a11a495843466cf80bd32f3bcee95e4 uconfig.sh
  * ex: set ro: */

uconfig.sh

@@ -850,6 +850,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src='.'
 ssizetype=int
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='4'
 startperl='#!perl'

uconfig64.sh

@@ -850,6 +850,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src='.'
 ssizetype=long
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='8'
 startperl='#!perl'

win32/config.gc

@@ -1073,6 +1073,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src=''
 ssizetype='int'
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='4'
 startperl='#!perl'

win32/config.vc

@@ -1072,6 +1072,8 @@ srand48_r_proto='0'
 srandom_r_proto='0'
 src=''
 ssizetype='int'
+st_dev_sign='1'
+st_dev_size='4'
 st_ino_sign='1'
 st_ino_size='4'
 startperl='#!perl'