httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <>
Subject Re: more solaris stuff
Date Mon, 28 Jul 1997 19:06:32 GMT
I'm not convinced it's worth our time to figure out the exact ordering
that works some of the time.  I did try a few other orderings not obvious
from my test program.  I got bad results all the time.  I didn't try doing
SO_REUSEADDR before the bind however. 

I've removed the solaris_hup patch, but didn't update the known bugs page


On Mon, 28 Jul 1997, Marc Slemko wrote:

> No.  You are still missing the SO_REUSEADDR _before_ the bind(),
> but after the socket().  Add that, and you just get the odd behavior
> of the socket just being in BOUND second time around, not in LISTEN
> like it should be.  ie. first time through, it will accept a
> connection.  Second time through, it will not give any errors but
> the socket will be in BOUND state not LISTEN and won't accept
> connections.

This seems worse than getting "Address in use".

> This is the same thing I was seeing with Apache under Solaris when I added
> lots of Listen directives at once for ports that had been bound to before
> by the same process and did a HUP.
> > We should never dup a socket under solaris.  Even with SIGUSR1 it's a
> I'm not convinced of that.  I will look at the test program further and
> perhaps bounce it off someone at Sun.

Well even if you reorder SO_REUSEADDR you can't restart the server right?


View raw message