Re: [v3] Fix second half of libstdc++/6745

[Jonathan Lennox <lennox at cs dot columbia dot edu>]

[gcc-patches removed from response, as requested.]
On Tuesday, November 19 2002, "Benjamin Kosnik" wrote to "Jonathan Lennox, pcarlini@unitus.it, gdr@integrable-solutions.net, libstdc++@gcc.gnu.org, gcc-patches@gcc.gnu.org" saying:

> >A potentially much simpler solution occured to me.  Could we declare that
> >Paolo's original patch (the one that started this thread) only applies to
> >cin?  I.e., make the condition this:
> >    if (__sbin == cin->rdbuf() && isatty(0))
> 
> Hey y'all. I think a better way to think about this is that the
> streambuf is unbuffered. That way you don't have to deal with the isatty
> issue. (For unbuffered streams, _M_buf_size_opt == 0)
> 
> If you do the checks based on that I think it will be cleaner. 

The only downside to this it slows down stream << cin.rdbuf() unnecessarily
when cin is not a tty but sync_with_stdio is true.  (Since that operation
"empties" stdin, you don't have to worry about synchronization with the C
APIs.)

The upside is that it allows user-defined types to set their preferred
buffering sizes, by calling setbuf().  It's also enormously cleaner and more
natural.

I think this is enough of a win that I agree this is the best way to go.  I
didn't have my tree up to date when I was browsing the code originally, so I
didn't see the 8399 fix, or I think I would have suggested this from the
beginning.

So that's
   __size = __sbin->_M_buf_size_opt > 0 ? __sbin->_M_buf_size_opt : 1;
in Paolo's original patch, just to avoid infinite loops for a zero-size
optimal buffer.

-- 
Jonathan Lennox
lennox at cs dot columbia dot edu