[Sbcl-devel] unread-char and echo-streams

[Richard M K. <kr...@pr...>]

Hello,

Should UNREAD-CHAR on an instance of ECHO-STREAM unread the character
onto the echo-stream's input stream?

I can't find anything in the CLHS to determine the issue, but it turns
out that SBCL and CMUCL, which don't unread onto the underlying input
stream, disagree with all of CCL, Clisp, Allegro, LispWorks, ECL, and
ABCL, which do propagate the unread; and our behavior for echo-streams
differs from our behavior for two-way streams, which also propagate the
unread.  The behavior everybody else implements has the nice consequence
that whatever character terminates a READ on an echo-stream is always
the next character available on the echo-stream's input stream.  Does
anybody object to changing our echo-streams to propagate the unread?

Thanks,
Richard