Re: possible deadlock on invalid UTF-8 data

[rsteinke w-link net]

From: Havoc Pennington <hp redhat com>
>
> Daniel Elstner <daniel elstner gmx net> writes:
> > 
> > the utf8_skip_data[] array (glib/gutf8.c:104) contains 0 at index 0xfe
> > and 0xff.  This could easily cause endless loops when iterating over a
> > UTF-8 string by using g_utf8_next_char().
> > 
> > I know that 0xfe and 0xff are forbidden in UTF-8 strings, but those
> > shouldn't cause a deadlock IMHO.  Sometimes it's just not appropriate to
> > validate every string before passing it to the g_utf8_* functions.
>
> If you use next_char on invalid UTF-8, it can easily skip onto invalid
> memory - so you have to validate first to be safe, even with the
> change you've suggested.
>
> The policy for GLib and GTK is that _all_ UTF-8 must be validated, and
> that none of the functions are safe against invalid UTF-8, with a few
> specific exceptions (the GMarkup parser is safe, and g_utf8_validate()
> itself is obviously safe).

And there are a few other weird ones. For example, g_io_channel_write_chars()
requires valid UTF-8, but you're allowed to write partial characters
(have sequential writes, with the breaks between writes not falling
on character boundaries).

Ron Steinke