I have the following code

*sql_execute
PARAMETERS tnConnection,tcSql,tcOutput

IF EMPTY(tcSql)  && Processing via an SQLPREPARE() statement
	tcSql = " "   && So Error Routine has character value
	lnResult = SQLEXEC(tnConnection)
ELSE
	IF EMPTY(tcOutput)
		lnResult = SQLEXEC(tnConnection,tcSql)
	ELSE
		lnResult = SQLEXEC(tnConnection,tcSql,tcOutput)
	ENDIF
ENDIF

IF lnResult = -1 
	=AERROR(aTmpError)
	lnStack = ASTACKINFO(aTmpStack)

	IF TYPE("app_global") <> "O"
		app_global = CREATEOBJECT('custom')
		app_global.addproperty("app_name","Insert App Name")
		app_global.addproperty("employee","RPUPKO")
	ENDIF
		
	INSERT INTO spotlight_errors(application,sql_statement,error_msg,dt_error,userid,stack) VALUES;
	    (app_global.app_name,tcSql,aTmpError[2],DATETIME(),app_global.employee,this.array_to_char('aTmpStack'))
	USE IN (SELECT('spotlight_errors'))

	IF this.showerror
		=MESSAGEBOX(aTmpError[2],0,PROGRAM(aTmpStack[ALEN(aTmpStack,1)-1,1]))
		SET STEP ON 
	ENDIF
	RETURN .F.
ENDIF

One of the records in the error log has the error message Variable 'LNRESULT' is not found."

I've never been able to reproduce this error and it's quite possible that the environment would actually absorb it and allow processing to continue. So if this error were generated by IF lnResult = -1 the record would actually get written.

What I don't understand is how the variable could not have a value. Can anybody give me a set of circumstances where this variable does not exist when AERROR() is called?

One possibility is that the error was actually generated in some other module and properly handled there. It could have been the last error generated so that's what AERROR() would find. In which case, for this module it really has a value of -1 so I'm supposed to go into the error handling portion of the code. The problem with that is the value is being assigned by the SQLEXEC() statements. If that statement resulted in a value of -1, why doesn't AERROR() return the ODBC errors I'm expecting.

I've checked with the users of the system and nobody has noticed any unexpected results, so it appears that the final processing is correct. But this is driving me crazy(ier). Any and all suggestions gratefully accepted.

..........Rich