Re: fflush & close behavior not well-defined

[Andrew J. Schorr]

Hi,

On Sun, Sep 13, 2020 at 11:50:17PM -0600, arnold@skeeve.com wrote:
> Thanks for the report. Gawk's close already takes an optional second
> argument ("from" or "to") for closing one end of a two-way pipe.

True, but if you look at the code in io.c:do_close, you'll see that the from/to
"how" argument does not come into play until after the open file is identified
by scanning the red_head list. So the presence of that argument does not limit
the search to the subset of open pipes. It merely triggers a warning from
io.c:close_redir if used with a file that's not a pipe. So in the case of
close, the fix would require up to 3 arguments:
   close(file, [, filetype [, how]])
(with the code being smart enough to detect when "how" is located in the
2nd position).

> If you're willing to hack on the code a bit, I'd have it search the
> full list, and upon finding a second instance of the same string,
> close it also, but with a warning message.

I considered that, but what then would the return value be?

> Otherwise, I think that:
> 
> > Or we could just punt and pretend that nobody ever noticed this.
> 
> Is the right answer.

Plausibly. So not worth documenting?

As you know, this relates to the other matter we were discussing about the
non-uniqueness of redirection strings and the more general requirement for a
filetype for correct identification. It seemed useful to provide a simple
example of how the current paradigm is slightly broken.

Regards,
Andy