Re: backward compatability of tools

[Bruce Korb]

"Dr. David Kirkby" wrote:
> 
> Hi,
>         I've found in the past bugs in software are often found on one
> platform that don't seem to get noticed on another. For this reason I
> just tried to build gcc-2.0 on an old Sun SPARCstation 20 running
> SunOs 4.1.4 (ie hardware/software about 10 years old) to test some
> software.

<< travails elided >>

> Given I thought the aim of autoconf/automake was to allow easy
> configuration of software, does one not think supporting older
> hardware useful? I'd certainly like to be able to configure my
> software so it runs on new and old hardware, if that is possible.

Depends on:

1.  who you talk to
2.  whether or not you, personally, want to put the effort into it.

By adding code to accommodate machines that only a miniscule handful
of people use, maintainers have to sqander their precious time
messing with the extra code and everyone else must sit through
the accumulated fractions of seconds while tests are made to see
if the current, up-to-date platform is really one of the ancient ones.
Bad trade off all the way around.  A useful goal is to be able to
target reasonably current versions of HP and IBM and Sun and Linux et al.