Re: [Help-bash] catching read error conditions

[Mart Frauenlob]

On 29.01.2013 01:51, Dan Douglas wrote:
> On Monday, January 28, 2013 06:27:24 PM Mart Frauenlob wrote:
> > On 28.01.2013 17:48, Chet Ramey wrote:
> > > On 1/28/13 11:43 AM, Mart Frauenlob wrote:
...]
> > > > (($? > 128)) && printf "timeout reached, or signal received\n"
> > > >
> > > > will that be correct and generic enough, if I write a bash-only script?
> > >
> > > Yes, that would work.
[...]
> Also, for Bash's read:
>
>    /* `read -t 0 var' tests whether input is available with select/FIONREAD,
>       and fails if those are unavailable */
>    if (have_timeout && tmsec == 0 && tmusec == 0)
> #if 0
>      return (EXECUTION_FAILURE);
> #else
>      return (input_avail (fd) ? EXECUTION_SUCCESS : EXECUTION_FAILURE);
> #endif
>
> ...
>
> Don't know whether that's worth testing or improves reliability in your case.

Thank you for noting that special case.
So if no fd is available read returns failure (anyways?),
if TMOUT=0 and and fd is available, read will wait for input "forever", 
right?
oh, are TMOUT and the -t option of read fully equal in effect?
It looks like to me, but afaics the man page does not explicitly say so.


P.S. some fun for inexperienced colleges :) (to put into startup file?):

trap 'printf "you are too lazy\n";exit' EXIT
TMOUT=1