Closing standard I/O streams ruins SBCL

[Roman Marynchak]

I know that it is unusual and funny, but anyway try both cases:

(close *standard-input*)

and

(close *standard-output*)

in REPL. On Win32 this leads to the more serious problems (infinite loop with printing VirtualAlloc info) than on Linux, but even on Linux the handling is worse than in CLISP (CLISP recovers from this by doing RESET).

Also, on Linux *standard-output* seems to be reopened after the abort, but *standard-input* is not.

This is low-priority, of course.

Regards,
Roman