Almost impossible to debug when using (load NOMESSAGE)

[Järneström Jonas]

In GNU Emacs 20.7.1 (sparc-sun-solaris2.6, X toolkit)
 of Fri Mar 15 2002 on balmer
configured using `configure  --prefix=/tools/gnu/emacs/20.7 --with-pop 
--with-kerberos --with-x-toolkit=lucid'

When executing my elisp library today I got a 
"file mode specification error: end-of-file" message in the echo area.

When examining the *Messages* buffer, I wrongly believed that the
offending file was the one just above the error message (which was
mirrored in the *Messages* buffer).

But it wasn't. The offending file was a file loaded with (load
NOMESSAGE), which suppresses the loading message in the echo area AS
WELL as to the *Messages* buffer. The net result was that there was NO
TRACE AT ALL of the load operation that failed, which mislead me
into suspecting the wrong file.

I think this is very tricky and deceptive. Had this error occurred for
and end-user of my package on another site, and he had bug-reported it 
to me, I would never have been able to track it down, without physically
visiting his machine and checking his locally generated elisp files,
and even then it would have been hard to track down.

The reason I used (load NOMESSAGE) is that I dont want to bother the
end-user with a heap of load messages that is only disturbing for him.
Now realizing how NOMESSAGE works, there's NO WAY I dare to use
NOMESSAGE for my own sake as a maintainer of code that I'm going to
distribute to professional customers.

My change request is that NOMESSAGE should only affect the echo area,
i.e. NOT suppress the trace printing to the *Messages* buffer.
The current behaviour of NOMESSAGE is, in my view, very unfortunate
and potentially malicious.