Small inconsistencies in configure checks

[Olly Betts]

On Sat, Dec 02, 2023 at 10:24:57PM +0300, Vitaly Chikunov wrote:
> JFYI. I am not sure it's worth reporting but, when building 2.5.5 for ALT

Talking of small inconsistencies, 2.5.5 isn't a Xapian version...

> I noticed small inconsistencies in configure output.
> 
> 1. xapian-binding:
> 
>   checking for /usr/bin/rdoc... no
>   checking for rdoc... /usr/bin/rdoc
> 
> Looks curious but no problem since it's found anyway.

I think this is a bug but it's likely only a problem if you're
overriding RUBY=/path/to/ruby but not overriding RDOC as well.

The aim is to try to use the rdoc which matches the Ruby interpreter
found as a default (unless the user has explicitly overridden by
running ./configure RDOC=/path/to/rdoc or similar), but it looks like
the way that's currently being done doesn't actually work.  I'll fix
that.

> 2. xapian-core when built with GCC:
> 
>   checking whether __builtin_add_overflow is declared... yes
>   ...
>   checking whether _addcarry_u64 is declared... no
> 
> There is actually _addcarry_u64 in GCC too, but it's in x86intrin.h
> instead of intrin.h. This is no problem either, since it's superimposed
> by __builtin_add_overflow anyway.

I think __builtin_add_overflow() is going to be as good an option (and
possibly a better option where both are supported) so this doesn't seem
a problem unless there's actually some platform with x86intrin.h but
without both __builtin_add_overflow() and intrin.h.

Cheers,
    Olly