Обсуждение: autoconf check for AF_UNIX sockets

Hello,

can our autoconf guru create a test for checking the availability of AF_UNIX
sockets? It could be defined in config.h as HAVE_UNIX_SOCKET or similar. It
will enable to use them in the newest cygwin where are this sockets
implemented.
    Dan

----------------------------------------------
Daniel Horak
network and system administrator
e-mail: horak@sit.plzen-city.cz
privat e-mail: dan.horak@email.cz ICQ:36448176
----------------------------------------------

Horák Daniel writes:

> can our autoconf guru create a test for checking the availability of AF_UNIX
> sockets? It could be defined in config.h as HAVE_UNIX_SOCKET or similar. It
> will enable to use them in the newest cygwin where are this sockets
> implemented.

I'll check into it. (No presumptions about guru status made... :-) )

-- 
Peter Eisentraut                  Sernanders väg 10:115
peter_e@gmx.net                   75262 Uppsala
http://yi.org/peter-e/            Sweden

I wrote:

> > can our autoconf guru create a test for checking the availability of AF_UNIX
> > sockets? It could be defined in config.h as HAVE_UNIX_SOCKET or similar. It
> > will enable to use them in the newest cygwin where are this sockets
> > implemented.
> 
> I'll check into it.

A classical Autoconf test is impractical. First of all there's no reliable
compile-time evidence regarding these Unix sockets so we'd have to run a
program from configure. That's already a semi-no-no because it will break
cross-compilation and it also sounds a bit like a security concern.
Moreover, it's still doubtful whether you could learn a lot this way,
perhaps the user that runs configure cannot create these sockets or not
where configure is trying to create it, etc.

I have added a HAVE_UNIX_SOCKETS symbol into config.h.in that currently
checks !defined(__CYGWIN__) && !defined(__QNX__) in the accustomed manner.
You could extend it with specific Cygwin version checks.


-- 
Peter Eisentraut                  Sernanders väg 10:115
peter_e@gmx.net                   75262 Uppsala
http://yi.org/peter-e/            Sweden

On Sun, 20 Aug 2000, Peter Eisentraut wrote:

> I wrote:
> 
> > > can our autoconf guru create a test for checking the availability of AF_UNIX
> > > sockets? It could be defined in config.h as HAVE_UNIX_SOCKET or similar. It
> > > will enable to use them in the newest cygwin where are this sockets
> > > implemented.
> > 
> > I'll check into it.
> 
> A classical Autoconf test is impractical. First of all there's no reliable
> compile-time evidence regarding these Unix sockets so we'd have to run a
> program from configure. That's already a semi-no-no because it will break
> cross-compilation and it also sounds a bit like a security concern.
> Moreover, it's still doubtful whether you could learn a lot this way,
> perhaps the user that runs configure cannot create these sockets or not
> where configure is trying to create it, etc.

can't you just do a link test that checks that AF_UNIX is defined?

> > can't you just do a link test that checks that AF_UNIX is defined?
> 
> That doesn't say anything about whether the Unix sockets 
> really work, as
> systems where they don't work define this occasionally.

My idea was to create a test that will try to compile, link and run a small
program. The program should contain "int s = socket(AF_UNIX, ...) and a test
for succesful creation of 's'. Only when this program will compile and run
OK then it should be defined HAVE_AF_UNIX_SOCKET.
    Dan