Fabio,

It looks like a bug to me too.
Another observation. According to ICASE() help the following 2 statements are similar but the second one generates an error. They both generate an error w/o COUNT()

SELECT COUNT(ICASE(.F.,1)) FROM RUNROWS INTO CURSOR c2
SELECT COUNT(ICASE(.F.,1, Null)) FROM RUNROWS INTO CURSOR c2

>Expected: same result
>
>Observed: when eOtherwiseResult is absent,
>VFP return a Table of 1.
>

>CLEAR
>
>CREATE CURSOR RUNROWS (F1 L)
>FOR K=1 TO 500
>	APPEND BLANK
>NEXT
>? "ICASE(.T.,1,NULL)=ICASE(.T.,1)",ICASE(.T.,1,NULL)=ICASE(.T.,1)
>SYS(3054,12)
>SELECT COUNT(ICASE(.T.,1,NULL)) FROM RUNROWS
>? _TALLY
>SELECT COUNT(ICASE(.T.,1)) FROM RUNROWS
>? _TALLY
>SYS(3054,0)
>

>
>Of course this bug is present on all the grouping functions:
>

>MIN(),MAX()
>SUM(),AVG() important !!!
>