Re: Minimal working example (Re: set_buffer_mode)

> On the topic of configure: who had the brilliant idea to run
> configure for every module/directory separately for all the basic
> tests?  One would think that all basic properties can be determined
> once.

They can, sort of, if the submodule tests for compiler etc are removed
everything still works perfectly. I was looking at exactly that for
pike 8.1.

Also, some modules like 'system' just has an insane amount of tests
that are redundant.

If there is a getpwent_r there is bound to also be getpwent, etc.
Also, we only use one.

The tests for 'alternative' functions should really be done like so:

AC_HAVE_FUNCS(getpwent_r getpwent, [break;])

since we only need one of them. There are a lot of those.

We should also really test for things like C99 and various posix
functions, and then just use things that are guaranteed by those
standards without tets.

But all this requires a lot of rewriting of configure scripts, which
is anything but fun.

On a related note it is somewhat pointless to look for things in a
library that is known to not exist, all tests should be conditional on
the library existing at all.

-- 
Per Hedbor