Created attachment 13956 [details] [review]
gspawn.patch

er, doh. type-o in my patch

Created attachment 13957 [details] [review]
gspawn.patch w/o the type-o :-)

- Did this fix the problem you were running into? 

- To be _real_ picky, I'm not that fond of the double
  loop in your full_write() function. My idea of "best
  practices" for this is somethign like write_all()
  gdk/x11/linux-fb/gdkmouse-fb.c:

==
static gboolean
write_all (gint   fd,
           gchar *buf,
           gsize  to_write)
{
  while (to_write > 0)
    {
      gssize count = write (fd, buf, to_write);
      if (count < 0)
        {
          if (errno != EINTR)
            return FALSE;
        }
      else
        {
          to_write -= count;
          buf += count;
        }
    }

  return TRUE;
}
==

Created attachment 13962 [details] [review]
new gspawn.patch using write_all()

*poke*

this should go into both 2.0.x and 2.2.x preferably

You indicated earlier that this doesn't actually fix any
reproducible bugs, so I didn't figure there was urgency
to get to it before we next worked on a 2.2.x release.

2.0.x is dead, as announced, no future releases will
be made off the gtk-2-0 branch. (Did you mean 
2.2 and head?)

well, I meant 2.0 and 2.2 - but since 2.0 is dead, I guess 2.2 and
head, yea :)

I suppose there's no urgency

Applied to both branches with the tiny change of using
glib types in the write_all() prototype.

looking at it, it's a bit weird that we have different
styles and naming for read_data() and write_all(), but not
weird enough that I'm going to bother changing things.