Re: [PATCH] tee: Add --pipe-check to allow instantly detecting closed outputs

[Carl Edquist]

On Mon, 12 Dec 2022, Arsen Arsenović wrote:

> Hi Rob,
>
> Rob Landley <rob@landley.net> writes:
>
> > This sort of thing is why I added -i to toybox's "timeout" command:
> >
> >     -i  Only kill for inactivity (restart timeout when command produces output)
> >
> > It runs the command's stdout through a pipe and does a poll() with the 
> > -i seconds value, and signals the program if the poll() expires.
> >
> > The android guys found it useful, but I was waiting to hear back about 
> > "cut -DF" before bringing it up here...
>
> That's interesting, might be worth adding to the GNU timeout, however, 
> it's not appropriate for what I'm using tee for, since compiler 
> processes could appear idle for a long time, if doing LTO for instance.

Thanks Rob for the idea.

Sounds like it could be useful if you know what an appropriate timeout 
ought to be ... though in the general data-driven case i feel like filters 
should only quit for relevant events like end-of-input or write failure. 
(The broken-pipe detection idea being an extension that detects the point 
when future writes will fail.)

Sounds like a neat idea to add to a timeout command (where you know the 
output timeout you want) if it can be made to work with arbitrary 
programs.

Carl