gcl, mformat and *error-output*

[Leo Butler]

(I posted this to maxima-discuss yesterday, but have not had a reaction,
so I am re-posting here. This bug can be reproduced with the current
Maxima+GCL build in Debian testing.)


I am trying to figure out what is causing a vexing bug involving GCL,
MFORMAT and *ERROR-OUTPUT*.

Consider a stripped-down definition of MY-ERROR-FN:

MAXIMA> (defun my-error-fn (sstring &rest l)
         (let ((*standard-output* *error-output*))
           (mformat t "~a" sstring)
           (fresh-line *standard-output*)))

The return value of MY-ERROR-FN is T if FRESH-LINE prints a newline and
NIL if not.

Here it is with the default value of *ERROR-OUTPUT*:

MAXIMA> (my-error-fn "An error occurred.")
An error occurred.
NIL

That is a bug, FRESH-LINE should have inserted a newline. What is weird
is that if MFORMAT is replaced by FORMAT, then FRESH-LINE works as
expected. I cannot figure out what MFORMAT is doing that tickles this
bug. Note that this is specific to GCL (tested with v2.6.14).

----

Here are two good cases, where FRESH-LINE works correctly.

MAXIMA> (setf *error-output* *standard-output* )

#<synonym stream to *TERMINAL-IO*>

MAXIMA> (my-error-fn "An error occurred.")
An error occurred.
T


MAXIMA> (setf *error-output* (make-string-output-stream ))

#<string-output stream 000000000257bc18>
MAXIMA> (my-error-fn "An error occurred.")

T
MAXIMA> (get-output-stream-string *error-output*)

"An error occurred.
"

TIA,
Leo