libdb2/test/run.test

#!/bin/sh -
#
#	@(#)run.test	8.13 (Berkeley) 11/2/95
#

. ./runenv.sh

# db regression tests
main()
{

	PROG=./dbtest
	TMP1=${TMPDIR-.}/t1
	TMP2=${TMPDIR-.}/t2
	TMP3=${TMPDIR-.}/t3
	BINFILES=${TMPDIR-.}/binfiles

	if [ \! -z "$WORDLIST" -a -f "$WORDLIST" ]; then
		DICT=$WORDLIST
	elif [ -f /usr/local/lib/dict/words ]; then
		DICT=/usr/local/lib/dict/words
	elif [ -f /usr/share/dict/words ]; then
		DICT=/usr/share/dict/words
	elif [ -f /usr/dict/words ]; then
		DICT=/usr/dict/words
	elif [ -f /usr/share/lib/dict/words ]; then
		DICT=/usr/share/lib/dict/words
	elif [ -f $srcdir/../test/dictionary ]; then
		DICT=`cd $srcdir/../test && pwd`/dictionary
	else
		echo 'run.test: no dictionary'
		exit 1
	fi

	dictsize=`wc -l < $DICT`

	bindir=/bin/.
	find $bindir -type f -exec test -r {} \; -print | head -100 > $BINFILES

	if [ $# -eq 0 ]; then
		for t in 1 2 3 4 5 6 7 8 9 10 11 12 13 20 40 41 50 60 61 62 63; do
			test$t
		done
	else
		while [ $# -gt 0 ]
			do case "$1" in
			test*)
				$1;;
			[0-9]*)
				test$1;;
			btree)
				for t in 1 2 3 7 8 9 10 12 13 40 41 50 60 61 62 63; do
					test$t
				done;;
			hash)
				for t in 1 2 3 8 13 20; do
					test$t
				done;;
			recno)
				for t in 1 2 3 4 5 6 7 10 11; do
					test$t
				done;;
			*)
				echo "run.test: unknown test $1"
				echo "usage: run.test test# | type"
				exit 1
			esac
			shift
		done
	fi
	rm -f $TMP1 $TMP2 $TMP3 $BINFILES
	exit 0
}

getnwords() {
	# Delete blank lines because the db code appears not to like
	# empty keys.  Omit lines with non-alphanumeric characters to
	# avoid shell metacharacters and non-ASCII characters which
	# could cause 'rev' to choke.
	LC_ALL=C sed -e '/^$/d' -e '/[^A-Za-z]/d' < $DICT | sed -e ${1}q
}

# Take the first hundred entries in the dictionary, and make them
# be key/data pairs.
test1()
{
	echo "Test 1: btree, hash: small key, small data pairs"
	getnwords 200 > $TMP1
	for type in btree hash; do
		rm -f $TMP2 $TMP3
		for i in `cat $TMP1`; do
			echo p
			echo k$i
			echo d$i
			echo g
			echo k$i
		done > $TMP2
		$PROG -o $TMP3 $type $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test1: type $type: failed"
			exit 1
		fi
	done
	echo "Test 1: recno: small key, small data pairs"
	rm -f $TMP2 $TMP3
	awk '{
		++i;
		printf("p\nk%d\nd%s\ng\nk%d\n", i, $0, i);
	}' < $TMP1 > $TMP2
	$PROG -o $TMP3 recno $TMP2
	if (cmp -s $TMP1 $TMP3) ; then :
	else
		echo "test1: type recno: failed"
		exit 1
	fi
}

# Take the first 200 entries in the dictionary, and give them
# each a medium size data entry.
test2()
{
	echo "Test 2: btree, hash: small key, medium data pairs"
	mdata=abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz
	echo $mdata |
	awk '{ for (i = 1; i < 201; ++i) print $0 }' > $TMP1
	for type in hash btree; do
		rm -f $TMP2 $TMP3
		for i in `getnwords 200`; do
			echo p
			echo k$i
			echo d$mdata
			echo g
			echo k$i
		done > $TMP2
		$PROG -o $TMP3 $type $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test2: type $type: failed"
			exit 1
		fi
	done
	echo "Test 2: recno: small key, medium data pairs"
	rm -f $TMP2 $TMP3
	echo $mdata |
	awk '{  for (i = 1; i < 201; ++i)
		printf("p\nk%d\nd%s\ng\nk%d\n", i, $0, i);
	}' > $TMP2
	$PROG -o $TMP3 recno $TMP2
	if (cmp -s $TMP1 $TMP3) ; then :
	else
		echo "test2: type recno: failed"
		exit 1
	fi
}

# Insert the programs in $bindir with their paths as their keys.
test3()
{
	echo "Test 3: hash: small key, big data pairs"
	rm -f $TMP1
	xargs cat < $BINFILES > $TMP1
	for type in hash; do
		rm -f $TMP2 $TMP3
		for i in `cat $BINFILES`; do
			echo p
			echo k$i
			echo D$i
			echo g
			echo k$i
		done > $TMP2
		$PROG -o $TMP3 $type $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test3: $type: failed"
			exit 1
		fi
	done
	echo "Test 3: btree: small key, big data pairs"
	for psize in 512 16384 65536; do
		echo "    page size $psize"
		for type in btree; do
			rm -f $TMP2 $TMP3
			for i in `cat $BINFILES`; do
				echo p
				echo k$i
				echo D$i
				echo g
				echo k$i
			done > $TMP2
			$PROG -i psize=$psize -o $TMP3 $type $TMP2
			if (cmp -s $TMP1 $TMP3) ; then :
			else
				echo "test3: $type: page size $psize: failed"
				exit 1
			fi
		done
	done
	echo "Test 3: recno: big data pairs"
	rm -f $TMP2 $TMP3
	awk '{
		++i;
		printf("p\nk%d\nD%s\ng\nk%d\n", i, $0, i);
	}' < $BINFILES > $TMP2
	for psize in 512 16384 65536; do
		echo "    page size $psize"
		$PROG -i psize=$psize -o $TMP3 recno $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test3: recno: page size $psize: failed"
			exit 1
		fi
	done
}

# Do random recno entries.
test4()
{
	echo "Test 4: recno: random entries"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk '{
		for (i = 37; i <= 37 + 88 * 17; i += 17) {
			if (i % 41)
				s = substr($0, 1, i % 41);
			else
				s = substr($0, 1);
			printf("input key %d: %s\n", i, s);
		}
		for (i = 1; i <= 15; ++i) {
			if (i % 41)
				s = substr($0, 1, i % 41);
			else
				s = substr($0, 1);
			printf("input key %d: %s\n", i, s);
		}
		for (i = 19234; i <= 19234 + 61 * 27; i += 27) {
			if (i % 41)
				s = substr($0, 1, i % 41);
			else
				s = substr($0, 1);
			printf("input key %d: %s\n", i, s);
		}
		exit
	}' > $TMP1
	rm -f $TMP2 $TMP3
	cat $TMP1 |
	awk 'BEGIN {
			i = 37;
			incr = 17;
		}
		{
			printf("p\nk%d\nd%s\n", i, $0);
			if (i == 19234 + 61 * 27)
				exit;
			if (i == 37 + 88 * 17) {
				i = 1;
				incr = 1;
			} else if (i == 15) {
				i = 19234;
				incr = 27;
			} else
				i += incr;
		}
		END {
			for (i = 37; i <= 37 + 88 * 17; i += 17)
				printf("g\nk%d\n", i);
			for (i = 1; i <= 15; ++i)
				printf("g\nk%d\n", i);
			for (i = 19234; i <= 19234 + 61 * 27; i += 27)
				printf("g\nk%d\n", i);
		}' > $TMP2
	$PROG -o $TMP3 recno $TMP2
	if (cmp -s $TMP1 $TMP3) ; then :
	else
		echo "test4: type recno: failed"
		exit 1
	fi
}

# Do reverse order recno entries.
test5()
{
	echo "Test 5: recno: reverse order entries"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk ' {
		for (i = 1500; i; --i) {
			if (i % 34)
				s = substr($0, 1, i % 34);
			else
				s = substr($0, 1);
			printf("input key %d: %s\n", i, s);
		}
		exit;
	}' > $TMP1
	rm -f $TMP2 $TMP3
	cat $TMP1 |
	awk 'BEGIN {
			i = 1500;
		}
		{
			printf("p\nk%d\nd%s\n", i, $0);
			--i;
		}
		END {
			for (i = 1500; i; --i)
				printf("g\nk%d\n", i);
		}' > $TMP2
	$PROG -o $TMP3 recno $TMP2
	if (cmp -s $TMP1 $TMP3) ; then :
	else
		echo "test5: type recno: failed"
		exit 1
	fi
}

# Do alternating order recno entries.
test6()
{
	echo "Test 6: recno: alternating order entries"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk ' {
		for (i = 1; i < 1200; i += 2) {
			if (i % 34)
				s = substr($0, 1, i % 34);
			else
				s = substr($0, 1);
			printf("input key %d: %s\n", i, s);
		}
		for (i = 2; i < 1200; i += 2) {
			if (i % 34)
				s = substr($0, 1, i % 34);
			else
				s = substr($0, 1);
			printf("input key %d: %s\n", i, s);
		}
		exit;
	}' > $TMP1
	rm -f $TMP2 $TMP3
	cat $TMP1 |
	awk 'BEGIN {
			i = 1;
			even = 0;
		}
		{
			printf("p\nk%d\nd%s\n", i, $0);
			i += 2;
			if (i >= 1200) {
				if (even == 1)
					exit;
				even = 1;
				i = 2;
			}
		}
		END {
			for (i = 1; i < 1200; ++i)
				printf("g\nk%d\n", i);
		}' > $TMP2
	$PROG -o $TMP3 recno $TMP2
	sort -o $TMP1 $TMP1
	sort -o $TMP3 $TMP3
	if (cmp -s $TMP1 $TMP3) ; then :
	else
		echo "test6: type recno: failed"
		exit 1
	fi
}

# Delete cursor record
test7()
{
	echo "Test 7: btree, recno: delete cursor record"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk '{
		for (i = 1; i <= 120; ++i)
			printf("%05d: input key %d: %s\n", i, i, $0);
		printf("%05d: input key %d: %s\n", 120, 120, $0);
		printf("seq failed, no such key\n");
		printf("%05d: input key %d: %s\n", 1, 1, $0);
		printf("%05d: input key %d: %s\n", 2, 2, $0);
		exit;
	}' > $TMP1
	rm -f $TMP2 $TMP3

	for type in btree recno; do
		cat $TMP1 |
		awk '{
			if (i == 120)
				exit;
			printf("p\nk%d\nd%s\n", ++i, $0);
		}
		END {
			printf("fR_NEXT\n");
			for (i = 1; i <= 120; ++i)
				printf("s\n");
			printf("fR_CURSOR\ns\nk120\n");
			printf("r\n");
			printf("fR_NEXT\ns\n");
			printf("fR_CURSOR\ns\nk1\n");
			printf("r\n");
			printf("fR_FIRST\ns\n");
		}' > $TMP2
		$PROG -o $TMP3 recno $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test7: type $type: failed"
			exit 1
		fi
	done
}

# Make sure that overflow pages are reused.
test8()
{
	echo "Test 8: btree: repeated small key, big data pairs"
	rm -f $TMP1
	# /bin/csh is no longer ubiquitous - find a substitute
	# The test stores contents of a known file
	tfile=""
	for tp in /bin/csh /bin/ls /usr/bin/ls /bin/cat /usr/bin/cat; do
	    if [ "x$tfile" = "x" -a -f $tp ]; then
		tfile=$tp
	    fi
	done
	if [ "x$tfile" = "x" ]; then
	    echo "No suitable file for testing purposes"
	    exit 1
	fi
	echo "" |
	awk 'BEGIN {
		for (i = 1; i <= 10; ++i) {
			printf("p\nkkey1\nD/bin/sh\n");
			printf("p\nkkey2\nD'$tfile'\n");
			if (i % 8 == 0) {
				printf("c\nkkey2\nD'$tfile'\n");
				printf("c\nkkey1\nD/bin/sh\n");
				printf("e\t%d of 10 (comparison)\n", i);
			} else
				printf("e\t%d of 10             \n", i);
			printf("r\nkkey1\nr\nkkey2\n");
		}
	}' > $TMP1
	if $PROG btree $TMP1 ; then
	    true
	else
	    echo "test8: btree tests failed"
	    exit 1
	fi
#	$PROG hash $TMP1
	# No explicit test for success.
}

# Test btree duplicate keys
test9()
{
	echo "Test 9: btree: duplicate keys"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk '{
		for (i = 1; i <= 543; ++i)
			printf("%05d: input key %d: %s\n", i, i, $0);
		exit;
	}' > $TMP1
	rm -f $TMP2 $TMP3

	for type in btree; do
		cat $TMP1 |
		awk '{
			if (i++ % 2)
				printf("p\nkduplicatekey\nd%s\n", $0);
			else
				printf("p\nkunique%dkey\nd%s\n", i, $0);
		}
		END {
				printf("o\n");
		}' > $TMP2
		$PROG -iflags=1 -o $TMP3 $type $TMP2
		sort -o $TMP3 $TMP3
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test9: type $type: failed"
			exit 1
		fi
	done
}

# Test use of cursor flags without initialization
test10()
{
	echo "Test 10: btree, recno: test cursor flag use"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk '{
		for (i = 1; i <= 20; ++i)
			printf("%05d: input key %d: %s\n", i, i, $0);
		exit;
	}' > $TMP1
	rm -f $TMP2 $TMP3

	# Test that R_CURSOR doesn't succeed before cursor initialized
	for type in btree recno; do
		cat $TMP1 |
		awk '{
			if (i == 10)
				exit;
			printf("p\nk%d\nd%s\n", ++i, $0);
		}
		END {
			printf("fR_CURSOR\nr\n");
			printf("eR_CURSOR SHOULD HAVE FAILED\n");
		}' > $TMP2
		$PROG -o $TMP3 $type $TMP2 > /dev/null 2>&1
		if [ -s $TMP3 ] ; then
			echo "Test 10: delete: R_CURSOR SHOULD HAVE FAILED"
			exit 1
		fi
	done
	for type in btree recno; do
		cat $TMP1 |
		awk '{
			if (i == 10)
				exit;
			printf("p\nk%d\nd%s\n", ++i, $0);
		}
		END {
			printf("fR_CURSOR\np\nk1\ndsome data\n");
			printf("eR_CURSOR SHOULD HAVE FAILED\n");
		}' > $TMP2
		$PROG -o $TMP3 $type $TMP2 > /dev/null 2>&1
		if [ -s $TMP3 ] ; then
			echo "Test 10: put: R_CURSOR SHOULD HAVE FAILED"
			exit 1
		fi
	done
}

# Test insert in reverse order.
test11()
{
	echo "Test 11: recno: reverse order insert"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk '{
		for (i = 1; i <= 779; ++i)
			printf("%05d: input key %d: %s\n", i, i, $0);
		exit;
	}' > $TMP1
	rm -f $TMP2 $TMP3

	for type in recno; do
		cat $TMP1 |
		awk '{
			if (i == 0) {
				i = 1;
				printf("p\nk1\nd%s\n", $0);
				printf("%s\n", "fR_IBEFORE");
			} else
				printf("p\nk1\nd%s\n", $0);
		}
		END {
				printf("or\n");
		}' > $TMP2
		$PROG -o $TMP3 $type $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test11: type $type: failed"
			exit 1
		fi
	done
}

# Take the first 20000 entries in the dictionary, reverse them, and give
# them each a small size data entry.  Use a small page size to make sure
# the btree split code gets hammered.
test12()
{
	if ( rev < /dev/null ) > /dev/null 2>&1 ; then
		:
	else
		echo "Test 12: skipped, rev not found"
		return
	fi
	if test $dictsize -lt 20001 ; then
		echo "Test 12: skipped, dictionary too small"
		return
	else
		:
	fi
	echo "Test 12: btree: lots of keys, small page size"
	mdata=abcdefghijklmnopqrstuvwxy
	echo $mdata |
	awk '{ for (i = 1; i < 20001; ++i) print $0 }' > $TMP1
	for type in btree; do
		rm -f $TMP2 $TMP3
		for i in `getnwords 20000 | rev`; do
			echo p
			echo k$i
			echo d$mdata
			echo g
			echo k$i
		done > $TMP2
		$PROG -i psize=512 -o $TMP3 $type $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test12: type $type: failed"
			exit 1
		fi
	done
}

# Test different byte orders.
test13()
{
	echo "Test 13: btree, hash: differing byte orders"
	getnwords 50 > $TMP1
	for order in 1234 4321; do
		for type in btree hash; do
			rm -f byte.file $TMP2 $TMP3
			for i in `cat $TMP1`; do
				echo p
				echo k$i
				echo d$i
				echo S
				echo g
				echo k$i
			done > $TMP2
			$PROG -ilorder=$order -f byte.file -o $TMP3 $type $TMP2
			if (cmp -s $TMP1 $TMP3) ; then :
			else
				echo "test13: $type/$order put failed"
				exit 1
			fi
			for i in `cat $TMP1`; do
				echo g
				echo k$i
			done > $TMP2
			$PROG -s \
			    -ilorder=$order -f byte.file -o $TMP3 $type $TMP2
			if (cmp -s $TMP1 $TMP3) ; then :
			else
				echo "test13: $type/$order get failed"
				exit 1
			fi
		done
	done
	rm -f byte.file
}

# Try a variety of bucketsizes and fill factors for hashing
test20()
{
	if test $dictsize -lt 10001 ; then
		echo "Test 20: skipped, dictionary too small"
		return
	else
		:
	fi
	echo\
    "Test 20: hash: bucketsize, fill factor; nelem 25000 cachesize 65536"
	echo "abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg" |
	awk '{
		for (i = 1; i <= 10000; ++i) {
			if (i % 34)
				s = substr($0, 1, i % 34);
			else
				s = substr($0, 1);
			printf("%s\n", s);
		}
		exit;
	}' > $TMP1
	getnwords 10000 |
	awk 'BEGIN {
		ds="abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg abcdefg"
	}
	{
		if (++i % 34)
			s = substr(ds, 1, i % 34);
		else
			s = substr(ds, 1);
		printf("p\nk%s\nd%s\n", $0, s);
	}' > $TMP2
	getnwords 10000 |
	awk '{
		++i;
		printf("g\nk%s\n", $0);
	}' >> $TMP2
	bsize=256
	for ffactor in 11 14 21; do
		echo "    bucketsize $bsize, fill factor $ffactor"
		$PROG -o$TMP3 \
		    -ibsize=$bsize,ffactor=$ffactor,nelem=25000,cachesize=65536\
		    hash $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test20: type hash:\
bsize=$bsize ffactor=$ffactor nelem=25000 cachesize=65536 failed"
			exit 1
		fi
	done
	bsize=512
	for ffactor in 21 28 43; do
		echo "    bucketsize $bsize, fill factor $ffactor"
		$PROG -o$TMP3 \
		    -ibsize=$bsize,ffactor=$ffactor,nelem=25000,cachesize=65536\
		    hash $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test20: type hash:\
bsize=$bsize ffactor=$ffactor nelem=25000 cachesize=65536 failed"
			exit 1
		fi
	done
	bsize=1024
	for ffactor in 43 57 85; do
		echo "    bucketsize $bsize, fill factor $ffactor"
		$PROG -o$TMP3 \
		    -ibsize=$bsize,ffactor=$ffactor,nelem=25000,cachesize=65536\
		    hash $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test20: type hash:\
bsize=$bsize ffactor=$ffactor nelem=25000 cachesize=65536 failed"
			exit 1
		fi
	done
	bsize=2048
	for ffactor in 85 114 171; do
		echo "    bucketsize $bsize, fill factor $ffactor"
		$PROG -o$TMP3 \
		    -ibsize=$bsize,ffactor=$ffactor,nelem=25000,cachesize=65536\
		    hash $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test20: type hash:\
bsize=$bsize ffactor=$ffactor nelem=25000 cachesize=65536 failed"
			exit 1
		fi
	done
	bsize=4096
	for ffactor in 171 228 341; do
		echo "    bucketsize $bsize, fill factor $ffactor"
		$PROG -o$TMP3 \
		    -ibsize=$bsize,ffactor=$ffactor,nelem=25000,cachesize=65536\
		    hash $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test20: type hash:\
bsize=$bsize ffactor=$ffactor nelem=25000 cachesize=65536 failed"
			exit 1
		fi
	done
	bsize=8192
	for ffactor in 341 455 683; do
		echo "    bucketsize $bsize, fill factor $ffactor"
		$PROG -o$TMP3 \
		    -ibsize=$bsize,ffactor=$ffactor,nelem=25000,cachesize=65536\
		    hash $TMP2
		if (cmp -s $TMP1 $TMP3) ; then :
		else
			echo "test20: type hash:\
bsize=$bsize ffactor=$ffactor nelem=25000 cachesize=65536 failed"
			exit 1
		fi
	done
}

# Test for a weird page split condition where an insertion into index
# 0 of a page that would cause the new item to be the only item on the
# left page results in index 0 of the right page being erroneously
# skipped; this only happens with one particular key+data length for
# each page size.
test40 () {
	echo "Test 40: btree: page split on index 0"
	e=:
	for psize in 512 1024 2048 4096 8192; do
		echo "    page size $psize"
		kdsizes=`awk 'BEGIN {
			psize = '$psize'; hsize = int(psize/2);
			for (kdsize = hsize-40; kdsize <= hsize; kdsize++) {
				print kdsize;
			}
		}' /dev/null`

		# Use a series of keylen+datalen values in the right
		# neighborhood to find the one that triggers the bug.
		# We could compute the exact size that triggers the
		# bug but this additional fuzz may be useful.

		# Insert keys in reverse order to maximize the chances
		# for a split on index 0.

		for kdsize in $kdsizes; do
			awk 'BEGIN {
				kdsize = '$kdsize';
				for (i = 8; i-- > 0; ) {
					s = sprintf("a%03d:%09d", i, kdsize);
					for (j = 0; j < kdsize-20; j++) {
						s = s "x";
					}
					printf("p\nka%03d\nd%s\n", i, s);
				}
				print "o";
			}' /dev/null > $TMP2
			sed -n 's/^d//p' $TMP2 | sort > $TMP1
			$PROG -o $TMP3 -i psize=$psize btree $TMP2
			if (cmp -s $TMP1 $TMP3); then :
			else
				echo "test40: btree: page size $psize, \
keylen+datalen=$kdsize failed"
				e='exit 1'
			fi
		done
	done
	$e
}

# Extremely tricky test attempting to replicate some unusual database
# corruption seen in the field: pieces of the database becoming
# inaccessible to random access, sequential access, or both.  The
# hypothesis is that at least some of these are triggered by the bug
# in page splits on index 0 with a particular exact keylen+datalen.
# (See Test 40.)  For psize=4096, this size is exactly 2024.

# The order of operations here relies on very specific knowledge of
# the internals of the btree access method in order to place records
# at specific offsets in a page and to create certain keys on internal
# pages.  The to-be-split page immediately prior to the bug-triggering
# split has the following properties:
#
# * is not the leftmost leaf page
# * key on the parent page is compares less than the key of the item
#   on index 0
# * triggering record's key also compares greater than the key on the
#   parent page

# Additionally, we prime the mpool LRU chain so that the head page on
# the chain has the following properties:
#
# * record at index 0 is located where it will not get overwritten by
#   items written to the right-hand page during the split
# * key of the record at index 0 compares less than the key of the
#   bug-triggering record

# If the page-split bug exists, this test appears to create a database
# where some records are inaccessible to a search, but still remain in
# the file and are accessible by sequential traversal.  At least one
# record gets duplicated out of sequence.

test41 () {
	echo "Test 41: btree: no unsearchables due to page split on index 0"
	# list of individual retrievals in a variable for easy reuse
	list=`(for i in a b c d; do
			for j in 990 998 999; do
				echo g ${i}${j} 1024
			done
		done;
		echo g y997 2014
		for i in y z; do
			for j in 998 999; do
				echo g ${i}${j} 1024
			done
		done)`
	# Exact number for trigger condition accounts for newlines
	# retained by dbtest with -ofile but not without; we use
	# -ofile, so count newlines.  keylen=5,datalen=5+2014 for
	# psize=4096 here.
	(cat - <<EOF
p z999 1024
p z998 1024
p y999 1024
p y990 1024
p d999 1024
p d990 1024
p c999 1024
p c990 1024
p b999 1024
p b990 1024
p a999 1024
p a990 1024
p y998 1024
r y990
p d998 1024
p d990 1024
p c998 1024
p c990 1024
p b998 1024
p b990 1024
p a998 1024
p a990 1024
p y997 2014
S
o
EOF
	echo "$list") |
	# awk script input:
	# {p|g|r} key [datasize]
	awk '/^[pgr]/{
		printf("%s\nk%s\n", $1, $2);
	}
	/^p/{
		s = $2;
		for (i = 0; i < $3; i++) {
			s = s "x";
		}
		printf("d%s\n", s);
	}
	!/^[pgr]/{
		print $0;
	}' > $TMP2
	(echo "$list"; echo "$list") | awk '{
		s = $2;
		for (i = 0; i < $3; i++) {
			s = s "x";
		}
		print s;
	}' > $TMP1
	$PROG -o $TMP3 -i psize=4096 btree $TMP2
	if (cmp -s $TMP1 $TMP3); then :
	else
		echo "test41: btree: failed"
		exit 1
	fi
}

# Test for recursive traversal successfully retrieving records that
# are inaccessible to normal sequential (sibling-link) traversal.
# This works by unlinking a few leaf pages but leaving their parent
# links intact.  To verify that the unlink actually makes records
# inaccessible, the test first uses "o" to do a normal sequential
# traversal, followed by "O" to do a recursive traversal.
test50 () {
	echo "Test 50: btree: recursive traversal"
	fill="abcdefghijklmnopqrstuvwxyzy"
	script='{
		for (i = 0; i < 20000; i++) {
			printf("p\nkAA%05d\nd%05d%s\n", i, i, $0);
		}
		print "u";
		print "u";
		print "u";
		print "u";
	}'
	(echo $fill | awk "$script"; echo o) > $TMP2
	echo $fill |
	awk '{
		for (i = 0; i < 20000; i++) {
			printf("%05d%s\n", i, $0);
		}
	}' > $TMP1
	$PROG -o $TMP3 -i psize=512 btree $TMP2
	if (cmp -s $TMP1 $TMP3); then
		echo "test50: btree: unexpected success after unlinking pages"
		exit 1
	fi
	(echo $fill | awk "$script"; echo O) > $TMP2
	$PROG -o $TMP3 -i psize=512 btree $TMP2
	if (cmp -s $TMP1 $TMP3); then :
	else
		echo "test50: btree: failed"
		exit 1
	fi
}

test60 () {
	echo "Test 60: btree: big key, small data, byteswap unaligned access"
	# 488 = 512 - 20 (header) - 3 ("foo") - 1 (newline)
	(echo foo; echo bar) |
	awk '{
		s = $0
		for (i = 0; i < 488; i++) {
			s = s "x";
		}
		printf("p\nk%s\ndx\n", s);
	}' > $TMP2
	for order in 1234 4321; do
		$PROG -o $TMP3 -i psize=512,lorder=$order btree $TMP2
	done
}

test61 () {
	echo "Test 61: btree: small key, big data, byteswap unaligned access"
	# 484 = 512 - 20 (header) - 7 ("foo1234") - 1 (newline)
	(echo foo1234; echo bar1234) |
	awk '{
		s = $0
		for (i = 0; i < 484; i++) {
			s = s "x";
		}
		printf("p\nk%s\nd%s\n", $0, s);
	}' > $TMP2
	for order in 1234 4321; do
		$PROG -o $TMP3 -i psize=512,lorder=$order btree $TMP2
	done
}

test62 () {
	echo "Test 62: btree: small key, big data, known byte order files"
	(echo foo1234; echo bar1234) |
	awk '{
		s = $0
		for (i = 0; i < 484; i++) {
			s = s "x";
		}
		printf("%s\n", s);
	}' > $TMP1
	(echo g; echo kfoo1234; echo g; echo kbar1234) > $TMP2
	for f in t.le.db t.be.db; do
		echo "    $f"
		$PROG -f $f -s -o $TMP3 btree $TMP2
		if (cmp -s $TMP1 $TMP3); then :
		else
			echo "test62: btree: failed"
			exit 1
		fi
	done
}

test63 () {
	echo "Test 63: btree: big key, medium data, bt_split unaligned access"
	# 488 = 512 - 20 (header) - 3 ("foo") - 1 (newline)
	# 223 = 232 - 8 (key pointer)
	# 232 = bt_ovflsize = (512 - 20 (header)) / 2 (DEFMINKEYPAGE)
	#	- (2 (indx_t) + 12 (NBLEAFDBT(0,0)))
	awk 'BEGIN {
		s = "";
		for (i = 0; i < 488; i++) {
			s = s "x";
		}
		d = "";
		for (i = 0; i < 223; i++) {
			d = d "x";
		}
		for (i = 0; i < 128; i++) {
			printf("p\nk%s%03d\nd%s\n", s, i, d);
		}
	}' /dev/null > $TMP2
	$PROG -o $TMP3 -i psize=512 btree $TMP2
}

main $*