>Fabio,
>
>They couild word it better but the SET TEXTMERGE TO MEMVAR scope continues until SET TEXTMERGE OFF. Any TRY...ENDTRY in that scope will reset memory variable when error occurs. Does it make sense to me? No, but it's the way it works.

Not. OFF don't close TEXTMERGE, it deactive the <<>> evaluation only.

CLEAR
LOCAL a
SET TEXTMERGE TO MEMVAR A
\1
SET TEXTMERGE OFF	&& OFF don't close the TEXTMERGE scope
\\2
A='0'
\\3
? "point 1",A
SET TEXTMERGE TO && MEMVAR A  -- THIS COPY THE TEXTMERGE'S BUFFER INTO THE VARIABLE
? "point 2",A
SET TEXTMERGE TO MEMVAR A ADDITIVE
\\5
SUBSCOPE()
\\6
SET TEXTMERGE TO && MEMVAR A  -- THIS COPY THE TEXTMERGE'S BUFFER INTO THE VARIABLE
? "point 4",A

PROCEDURE SUBSCOPE
	* A is out of scope, but the buffer is
	\\4
* uncomment next two lines
*	SET TEXTMERGE TO && MEMVAR A create a SUBSCOPE's private "a" variable
*	? "point 3",a

With a filename or _TEXT file handle, textmerge put every \|\\ line into the file directly;
with a memvar, TEXTMERGE build a text buffer, and only the
SET TEXTMERGW TO
copy/append the buffer into the variable.
If happen an error ( TRY/ENDTRY it is insignificant),
VFP executes a SET TEXTMERGE TO in the error's scope,
and therefore, the filename setting or the buffer value is lose.