>>
>>I'm not sure 100%, but this is correlate with the CHR(0)
>>( C++ nullString endtag ) and the index string COMPACT routine.
>>
>>If you define index with last key char = CHR(0) character, you can to meet this bug
>
>hi Fabio,
>
>I was thinking the other night there must be something in the key value that makes it fail. Dunno, maybe compact as you say or maybe it uses a bit of the key to indicate whether an index entry is occupied at certain places. I'm just guessing here.
>But if it does fail for a string( bintoc), then why would it not fail for an integer, which also returns a string ?

Gregory,
the bug is present on any string key with lenght > 1;
the COMPACT process is insignificant.

Next is a general repro

clear

	set collate to 'machine'


	cPrefix = 'someString' && any string with len>0
	
	N = 1  && 1..255
	
	create cursor CursorTest ;
		( 	j_ref		c(LEN(cPrefix)+1) ;
		)
	index on j_ref TAG j_ref

	* Start the index with  a fixed sequence
	FOR k=1 TO 2	&& COMPACT use the recno=1 key
	  insert into CursorTest values(m.cPrefix+CHR(m.N))  && prefix + not NULL ASCII
	NEXT
	* Append a Key with: same prefix, BUT END with NULL CHAR
	  insert into CursorTest values(m.cPrefix+CHR(0))	   
	? "FOUND",test() AT 20	&& now found it
	
	* Append another keys group with the keys equal to that of the start group
	FOR k=1 TO 10
	  insert into CursorTest values(m.cPrefix+CHR(m.N))  && prefix + not NULL ASCII
	  ? "NOT FOUND",test() AT 20
	NEXT
	* Append another key with a different key
	  insert into CursorTest values(m.cPrefix+CHR((m.N+1)%256))
	
	? "FOUND",test() AT 20	&& now the gost key it reappears

	
	PROCEDURE TEST
	  REINDEX
	  RETURN KEYMATCH(m.cPrefix+CHR(0))

For integers index VFP uses a different C++ routine,
and the bug is not present

clear

	set collate to 'machine'

	cPrefix = '340133' && any string
	
	N = 1  && 1..255
	
	create cursor CursorTest ;
		( 	j_ref		I ;
		)
	index on j_ref TAG j_ref

	insert into CursorTest values(0x01&cPrefix)  && prefix + not NULL ASCII
	
	* Append a Key with: same prefix, BUT END with NULL CHAR
	insert into CursorTest values(0x00&cPrefix)	   
	? "FOUND",test() AT 20	&& now found it
	
	* Append another keys group with the keys equal to that of the start group
	FOR k=1 TO 10
	insert into CursorTest values(0x01&cPrefix)  && prefix + not NULL ASCII
	? "NOT FOUND",test() AT 20
	NEXT
	* Append another key with a different key
	insert into CursorTest values(0x00&cPrefix)
	
	? "FOUND",test() AT 20	&& now the gost key it reappears
	
	PROCEDURE TEST
	REINDEX
	RETURN KEYMATCH(0x00&cPrefix)

The presence of the bug until version 8,
depends on the fact that is rare to have strings that they finish with CHR_0,
and nearly nobody uses BINTOC.
In VFP9, BINTOC has been upgraded
(still lacks the LOGICAL, the DATES and the DATETIME support),
and therefore the bug would have appeared a lot soon;
however now is resolved.

Still the master rule:

the bugs exists, also when they are not looked at


Fabio