Re: building groff with GNU Make and *BSD Make

[Larry McVoy]

On Tue, Mar 29, 2022 at 04:53:39PM +0200, Ingo Schwarze wrote:
> You seem to say "Pick *any* specific program, use the full feature
> set of one particular version of that specific program, and add a
> hard dependency on that specific version of that specific program."

No, I didn't say that, I said "pick one make, doesn't matter which one".
I guess you could take that to mean "pick one version of one make"
but I didn't say that and I did say that we had a few problems with
different versions of GNU make and, like you, backed off on using
whatever feature it was that didn't work on all versions of GNU make
we encountered.

> I say portability is an asset unless you have a good, specific
> reason to sacrifice it

I agree 100%, we have no argument there.  Where we seem to disagree
is whether it is worth the effort to be portable to every different
make implementation out there.  You seem to think that is good, and
in theory it is good, portable is good.  In practice, it is, in my
opinion, an opportunity cost.  It's free software, he's not getting
paid to work on it, so every problem has to be triaged, should this
get fixed or is it a "don't care"?

If there was infinite time, infinite resources, yep, portable is a
better answer.  There isn't infinite time, so then you have to make
a call.  We supported a system quite a bit larger and more complex
than groff, on pretty much the same set of platforms, and we quickly
decided that there needed to be one build system to rule them all.

Your mileage may vary.