David Noon wrote on 07 Jan 1996 at 12:25 to Ruud Senden:

RS> Well, I have tried some things, but it doesn't recognize the
RS> errors I  introduced.

 DN> Syntax checking will only be effective if your code is close to
 DN> valid. Anything that looks like complete gobbledy-gook to REXX will
 DN> be retained as is and passed to the ADDRESS handler that is active
 DN> at exxecution time.

I understand this, but I think it should see that a , is missing. 

 DN> However, a long REXX statement should burp at load time with a
 DN> comma missing. You wrote "REXX-command", so I am not sure whether
 DN> you mean an OS/2 command that REXX is to pass to CMD.EXE or a
 DN> native REXX statement. 

I meant REXX-statement, for example:

call wrt_msg 'Error changing directory to 'newDir'!' || nl || ,          
             'Please check the configuration file.'

When I leave the , away, REXX only sees the error when it comes to this statement.

 DN> I try to avoid issuing OS/2 commands when I am writing a REXX exec.
 DN> I find that using native REXX statements gives me much tighter
 DN> checking of my program's syntax, simply because it is all REXX
 DN> syntax. 

But for this REXX-script it is unavoidable, since the REXX-program is used
to automate a certain task (connecting to the Internet, exchanging mail and
news, and hanging up). I use REXX for this, so I can use very thight
error-checking (including timers), a configuration file (so other people
can easily configure it for their specific software) etc.

 DN> Of course, sometimes issuing a command is unavoidable, but
 DN> then I always use

 DN>   ADDRESS 'CMD' ''

 DN> so that it is brought back into REXX syntax for load-time checking.

Thank you for this good suggestion. Until now I always put an OS/2 command
on a line all by itself, so REXX would pass it to OS/2.

Regards,
  Ruud  (Internet: rsenden{at}cs.ruu.nl) 
--- timEd/2 1.10+