Re: Nonblocking get-bytevector-n bug?

[ludo@gnu.org (Ludovic Courtès)]

Marko Rauhamaa <marko@pacujo.net> skribis:

> Of these, only get-bytevector-some seems to behave as expected when the
> port is nonblocking:
>
> ========================================================================
> (use-modules (rnrs io ports)
>              (rnrs bytevectors))
>
> (let ((port (fdes->inport 0)))
>   (fcntl port F_SETFL (logior O_NONBLOCK (fcntl port F_GETFL)))
>   (let loop ()
>     (catch 'system-error
>       (lambda ()
>         (format #t "~s\n" (get-bytevector-some port)))
>       (lambda syserr
>         (format #t "err ~s\n" (system-error-errno syserr))))
>     (sleep 1)
>     (loop)))
> ========================================================================
>
> If you replace (get-bytevector-some port) with
> (get-bytevector-n port 8192) or (get-bytevector-n! port bv 0 8192),
> no partial data is returned.

That’s because get-bytevector-n{,!} are specified as blocking if
necessary until N bytes are read or EOF is reached.

> The problem with get-bytevector-some is that there is no limit to how
> many bytes might be returned. In practice, I see that the amount is
> capped at 4096 bytes, but the documentation does not guarantee any
> limit.

Right, the R6RS specification of that procedure doesn’t tell anything
more either.

It was criticized back in the day partly for that reason:

  http://www.r6rs.org/formal-comments/comment-224.txt

I think a ‘get-bytevector-some!’ procedure could address the issue.

Not sure what R7 does here.

Thoughts?

Ludo’.