Re: _AS_DETECT_BETTER_SHELL speedup

[Paul Eggert]

Stepan Kasal <address@hidden> writes:

> indeed, the speedup is not minor for Windows ports.
> For example, on Cygwin, each fork can take ages.

Let's see the actual measured performance numbers, for each proposed
change, on a modern Cygwin box.  The first rule for optimization that
hurts code quality is to avoid it if it doesn't really improve things
much overall.

> I don't think there is a risk of actually _breaking_ the configure
> script.

If the configure script outputs random irrelevant messages to users
then that will cause real problems in practice.  We get too many bug
reports already; I don't want to be bothered by bug reports about
bogus "not found" messages.

> Nothing worse can happen than a few spurious 'not found' messages.

I'm not 100% sure of this, unfortunately.