Boost logo

Boost Users :

From: Jonathan Turkanis (turkanis_at_[hidden])
Date: 2008-02-18 01:23:00

Chad Walters wrote:
> A couple thoughts:
> 1. One of the first confusions that I naively had when I first
> approached the library was that classes that had close() methods
> did not necessarily model Closable. I am not sure if this means
> that “Closable” is a bad name and should be changed, but it might
> be worth calling out this distinction more pointedly in the
> documentation.

I think I've finally made all classes with close() methods Closable, as
of revision 43299. Let me know if you find any more. Here's how
close() now behaves on various types that previously weren't Closable:

  type semantics
  ---- ---------

  ifstream, ofstream, s.close()
  fstream, filebuf,
  stream, stream_buffer

  istringstream, s.str("")

  filtering_stream, s.pop()

> 2. Perhaps for filtering streams, it would be best to make sure that
> closing the stream would close all the filters and flush the
> device.. Would that solve this issue? Or is that already the
> current behavior?

That's an interesting idea. It might work most of the time, but the
semantics of flush() don't actually require a device to do anything, so
the problem would probably still show up once in a while.

> Chad

Best Regards,

Jonathan Turkanis

Boost-users list run by williamkempf at, kalb at, bjorn.karlsson at, gregod at, wekempf at