Re: [Bug-ed] ed EOF behavior

[Antonio Diaz Diaz]

Andrew Moore wrote:
> Restoring GNU ed-1.7's  get_tty_line() function to its GNU ed 0.2
> equivalent (patch below) exposes another bug, presumably in GNU
> ed-1.7's append_lines().  So while Andrea's patch does indeed appear
> to address the annoying behavior of EOF in GNU ed-1.7, without
> looking deeper, I wonder whether it may not be masking more problems?
[...]
> "Double EOF" is a bug.

OK, lets begin another round of bug hunting in ed. :-)

In a few days I'll release ed-1.8-rc1 with the "Double EOF" behavior 
removed. If this is masking any problems, we'll discover and fix them.


> A comment should also be added at the top of  get_tty_line() that seems 
> relevant to Andrea' s efforts:
>   /* Read stdin one character at a time to avoid i/o contention
>      with shell escapes invoked by nonterminal input, e.g.,
>      ed - <<EOF
>      !cat
>      hello, world
>      EOF */

This is interesting, but not the only reason to read stdin one character 
at a time. Long lines and null characters are also reasons to read stdin 
this way.


> Finally, equating the null character '\0' with (int) 0 is asking for trouble...

Why?

The C standard states clearly that:
1) An integer character constant has type int.
2) The value of the escape sequence '\0' is 0.

So they are equivalent.


Regards,
Antonio.