Re: bread() patch

["Roy T. Fielding" <fi...@kiwi.ICS.UCI.EDU>]

> I can't figure out what is the best way to break out an error from here - have
> get_client_block return -1?  That's not checked for either.  The best thing I
> can think of to do is log an error and stop reading input.

I think returning -1 and letting the caller deal with the error message is
the right thing to do.  mod_fastcgi seems to expect the -1, and mod_cgi
can be easily fixed to do so.

In any case, the suggested patch isn't right, since it isn't a close of
connection (that would be an EOF, returning 0) and it is checking the
wrong variable (len_to_read instead of len_read).  I'll work on a better one.

.....Roy