bug#58416: -u should only affect stdout, not stdin

[Andras Korn]

Hi,

I just straced a running sed(1) instance (4.8), and was amazed to see it read 
its input character by character.

I then remembered that I called it with '-u' because its output was being read 
by a program that needed it ASAP.

I can't see any benefit to reading stdin by the character, but am prepared to 
believe a use-case exists. However, I submit that not all use-cases that 
require unbuffered (or line-buffered) output also automatically need 
inefficient reading of the input.

Maybe add a new switch for the input part?

Thanks and best regards,

András

-- 
                    Klein bottle for rent; enquire within.