Re: Not so useless use of cat

[Chet Ramey]

On 9/18/14, 9:26 AM, arnold@skeeve.com wrote:
> Chet Ramey <chet.ramey@case.edu> wrote:
> 
>> Yes, on this one.  There are others; recall the `discussion' about
>> whether bash should choose between /dev/fd or FIFOs for process
>> substitution at runtime.
> 
> That's a tougher one. It's a question of how far back do you wish
> to continue supporting systems?
> 
> I'm finding that modern systems have pretty much the union of things
> that I need, and also that the older ones that don't simply aren't
> in use anymore.  E.g., do you still need to support SunOS 4.1.x? Ultrix?
> OSF/1? Irix?
> 
> I removed a lot of crufty code in gawk ~3 years ago to zero complaints.
> You might be able to do that in Bash (e.g., assume /dev/fd), especially
> if you do it a major release point (Bash 5.0).

The problem isn't old, unused code.  It's the fact that some of the
things bash wants to use are optional OS features.  If bash is built on
a machine, say FreeBSD, with /dev/fd available at configure time, that
binary will attempt to use /dev/fd forever.  This isn't such a problem
when you're building bash yourself from ports, but it does inconvenience
binary distributions that provide a lot of user-modifiable configuration
options.  The same build/distribution problem exists on Linux.

The specific issue in the /dev/fd-FIFO case is using a feature like /dev/fd
at a point where /dev/fd is not available (e.g., during system boot).

Chet
-- 
``The lyf so short, the craft so long to lerne.'' - Chaucer
                 ``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, ITS, CWRU    chet@case.edu    http://cnswww.cns.cwru.edu/~chet/