X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=t%2Fop%2Flfs.t;h=2652555281e1d555bf8c9b7c1cb326f293eeb4c5;hb=b0f2b690b4ba59b02c372a35658748cb0f31c38e;hp=2683a69b00df5348651355c0ecde8344aa7c7e40;hpb=e9a694fc69831683df9b559726b85119e227aa26;p=p5sagit%2Fp5-mst-13.2.git

diff --git a/t/op/lfs.t b/t/op/lfs.t
index 2683a69..2652555 100644
--- a/t/op/lfs.t
+++ b/t/op/lfs.t
@@ -3,35 +3,75 @@
 # If you modify/add tests here, remember to update also t/lib/syslfs.t.
 
 BEGIN {
-	# Don't bother if there are no quads.
-	eval { my $q = pack "q", 0 };
-	if ($@) {
-		print "1..0\n# no 64-bit types\n";
-		exit(0);
-	}
 	chdir 't' if -d 't';
-	unshift @INC, '../lib';
+	@INC = '../lib';
 	# Don't bother if there are no quad offsets.
 	require Config; import Config;
 	if ($Config{lseeksize} < 8) {
-		print "1..0\n# no 64-bit file offsets\n";
+		print "1..0 # Skip: no 64-bit file offsets\n";
 		exit(0);
 	}
 }
 
-sub bye {
+use strict;
+
+our @s;
+our $fail;
+
+sub zap {
     close(BIG);
-    unlink "big";
+    unlink("big");
+    unlink("big1");
+    unlink("big2");
+}
+
+sub bye {
+    zap();	
     exit(0);
 }
 
+my $explained;
+
+sub explain {
+    unless ($explained++) {
+	print <<EOM;
+#
+# If the lfs (large file support: large meaning larger than two
+# gigabytes) tests are skipped or fail, it may mean either that your
+# process (or process group) is not allowed to write large files
+# (resource limits) or that the file system (the network filesystem?)
+# you are running the tests on doesn't let your user/group have large
+# files (quota) or the filesystem simply doesn't support large files.
+# You may even need to reconfigure your kernel.  (This is all very
+# operating system and site-dependent.)
+#
+# Perl may still be able to support large files, once you have
+# such a process, enough quota, and such a (file) system.
+# It is just that the test failed now.
+#
+EOM
+    }
+    print "1..0 # Skip: @_\n" if @_;
+}
+
+$| = 1;
+
+print "# checking whether we have sparse files...\n";
+
 # Known have-nots.
-if ($^O eq 'win32' || $^O eq 'vms') {
-    print "1..0\n# no sparse files\n";
+if ($^O eq 'MSWin32' || $^O eq 'NetWare' || $^O eq 'VMS') {
+    print "1..0 # Skip: no sparse files in $^O\n";
     bye();
 }
 
-# Then try to deduce whether we have sparse files.
+# Known haves that have problems running this test
+# (for example because they do not support sparse files, like UNICOS)
+if ($^O eq 'unicos') {
+    print "1..0 # Skip: no sparse files in $^0, unable to test large files\n";
+    bye();
+}
+
+# Then try to heuristically deduce whether we have sparse files.
 
 # Let's not depend on Fcntl or any other extension.
 
@@ -42,98 +82,193 @@ my ($SEEK_SET, $SEEK_CUR, $SEEK_END) = (0, 1, 2);
 # consume less blocks than one megabyte (assuming nobody has
 # one megabyte blocks...)
 
-open(BIG, ">big") or do { warn "open failed: $!\n"; bye };
-binmode BIG;
-seek(BIG, 1_000_000, $SEEK_SET);
-print BIG "big";
-close(BIG);
-
-my @s;
-
-@s = stat("big");
-
-print "# @s\n";
-
-my $BLOCKSIZE = 512; # is this really correct everywhere?
-
-unless (@s == 13 &&
-	$s[7] == 1_000_003 &&
-	defined $s[12] &&
-        $BLOCKSIZE * $s[12] < 1_000_003) {
-    print "1..0\n# no sparse files?\n";
-    bye();
+open(BIG, ">big1") or
+    do { warn "open big1 failed: $!\n"; bye };
+binmode(BIG) or
+    do { warn "binmode big1 failed: $!\n"; bye };
+seek(BIG, 1_000_000, $SEEK_SET) or
+    do { warn "seek big1 failed: $!\n"; bye };
+print BIG "big" or
+    do { warn "print big1 failed: $!\n"; bye };
+close(BIG) or
+    do { warn "close big1 failed: $!\n"; bye };
+
+my @s1 = stat("big1");
+
+print "# s1 = @s1\n";
+
+open(BIG, ">big2") or
+    do { warn "open big2 failed: $!\n"; bye };
+binmode(BIG) or
+    do { warn "binmode big2 failed: $!\n"; bye };
+seek(BIG, 2_000_000, $SEEK_SET) or
+    do { warn "seek big2 failed; $!\n"; bye };
+print BIG "big" or
+    do { warn "print big2 failed; $!\n"; bye };
+close(BIG) or
+    do { warn "close big2 failed; $!\n"; bye };
+
+my @s2 = stat("big2");
+
+print "# s2 = @s2\n";
+
+zap();
+
+unless ($s1[7] == 1_000_003 && $s2[7] == 2_000_003 &&
+	$s1[11] == $s2[11] && $s1[12] == $s2[12]) {
+	print "1..0 # Skip: no sparse files?\n";
+	bye;
 }
 
+print "# we seem to have sparse files...\n";
+
 # By now we better be sure that we do have sparse files:
 # if we are not, the following will hog 5 gigabytes of disk.  Ooops.
+# This may fail by producing some signal; run in a subprocess first for safety
 
-print "1..8\n";
+$ENV{LC_ALL} = "C";
 
-my $fail = 0;
+my $r = system '../perl', '-e', <<'EOF';
+open(BIG, ">big");
+seek(BIG, 5_000_000_000, 0);
+print BIG "big";
+exit 0;
+EOF
 
 open(BIG, ">big") or do { warn "open failed: $!\n"; bye };
 binmode BIG;
-seek(BIG, 5_000_000_000, $SEEK_SET);
-print BIG "big";
-close BIG;
+if ($r or not seek(BIG, 5_000_000_000, $SEEK_SET)) {
+    my $err = $r ? 'signal '.($r & 0x7f) : $!;
+    explain("seeking past 2GB failed: $err");
+    bye();
+}
+
+# Either the print or (more likely, thanks to buffering) the close will
+# fail if there are are filesize limitations (process or fs).
+my $print = print BIG "big";
+print "# print failed: $!\n" unless $print;
+my $close = close BIG;
+print "# close failed: $!\n" unless $close;
+unless ($print && $close) {
+    if ($! =~/too large/i) {
+	explain("writing past 2GB failed: process limits?");
+    } elsif ($! =~ /quota/i) {
+	explain("filesystem quota limits?");
+    } else {
+	explain("error: $!");
+    }
+    bye();
+}
 
 @s = stat("big");
 
 print "# @s\n";
 
+unless ($s[7] == 5_000_000_003) {
+    explain("kernel/fs not configured to use large files?");
+    bye();
+}
+
 sub fail () {
     print "not ";
     $fail++;
 }
 
+sub offset ($$) {
+    my ($offset_will_be, $offset_want) = @_;
+    my $offset_is = eval $offset_will_be;
+    unless ($offset_is == $offset_want) {
+        print "# bad offset $offset_is, want $offset_want\n";
+	my ($offset_func) = ($offset_will_be =~ /^(\w+)/);
+	if (unpack("L", pack("L", $offset_want)) == $offset_is) {
+	    print "# 32-bit wraparound suspected in $offset_func() since\n";
+	    print "# $offset_want cast into 32 bits equals $offset_is.\n";
+	} elsif ($offset_want - unpack("L", pack("L", $offset_want)) - 1
+	         == $offset_is) {
+	    print "# 32-bit wraparound suspected in $offset_func() since\n";
+	    printf "# %s - unpack('L', pack('L', %s)) - 1 equals %s.\n",
+	        $offset_want,
+	        $offset_want,
+	        $offset_is;
+        }
+        fail;
+    }
+}
+
+print "1..17\n";
+
+$fail = 0;
+
 fail unless $s[7] == 5_000_000_003;	# exercizes pp_stat
 print "ok 1\n";
 
 fail unless -s "big" == 5_000_000_003;	# exercizes pp_ftsize
 print "ok 2\n";
 
+fail unless -e "big";
+print "ok 3\n";
+
+fail unless -f "big";
+print "ok 4\n";
+
 open(BIG, "big") or do { warn "open failed: $!\n"; bye };
 binmode BIG;
 
-seek(BIG, 4_500_000_000, $SEEK_SET);
+fail unless seek(BIG, 4_500_000_000, $SEEK_SET);
+print "ok 5\n";
 
-fail unless tell(BIG) == 4_500_000_000;
-print "ok 3\n";
+offset('tell(BIG)', 4_500_000_000);
+print "ok 6\n";
 
-seek(BIG, 1, $SEEK_CUR);
+fail unless seek(BIG, 1, $SEEK_CUR);
+print "ok 7\n";
 
-fail unless tell(BIG) == 4_500_000_001;
-print "ok 4\n";
+# If you get 205_032_705 from here it means that
+# your tell() is returning 32-bit values since (I32)4_500_000_001
+# is exactly 205_032_705.
+offset('tell(BIG)', 4_500_000_001);
+print "ok 8\n";
 
-seek(BIG, -1, $SEEK_CUR);
+fail unless seek(BIG, -1, $SEEK_CUR);
+print "ok 9\n";
 
-fail unless tell(BIG) == 4_500_000_000;
-print "ok 5\n";
+offset('tell(BIG)', 4_500_000_000);
+print "ok 10\n";
 
-seek(BIG, -3, $SEEK_END);
+fail unless seek(BIG, -3, $SEEK_END);
+print "ok 11\n";
 
-fail unless tell(BIG) == 5_000_000_000;
-print "ok 6\n";
+offset('tell(BIG)', 5_000_000_000);
+print "ok 12\n";
 
 my $big;
 
 fail unless read(BIG, $big, 3) == 3;
-print "ok 7\n";
+print "ok 13\n";
 
 fail unless $big eq "big";
-print "ok 8\n";
+print "ok 14\n";
 
-if ($fail) {
-    print STDERR <<EOM;
-#
-# If the lfs (large file support) tests fail, it may mean that
-# the *file system* you are running the tests on doesn't support
-# large files (files larger than two gigabytes).  Perl may still
-# be able to support such files, once you have such a file system.
-#
-EOM
-}
+# 705_032_704 = (I32)5_000_000_000
+# See that we don't have "big" in the 705_... spot:
+# that would mean that we have a wraparound.
+fail unless seek(BIG, 705_032_704, $SEEK_SET);
+print "ok 15\n";
+
+my $zero;
+
+fail unless read(BIG, $zero, 3) == 3;
+print "ok 16\n";
 
-bye();
+fail unless $zero eq "\0\0\0";
+print "ok 17\n";
+
+explain() if $fail;
+
+bye(); # does the necessary cleanup
+
+END {
+   unlink "big"; # be paranoid about leaving 5 gig files lying around
+}
 
 # eof