FreeLists / openbeosnetteam / [openbeosnetteam] Re: Fix for resolving host names

[Scott Mansfield <thephantom@xxxxxxx>]

On Feb 2, 2004, at 8:39 AM, Brennan Cleveland wrote:

> I'll volunteer to work on libnet.so.  I am testing anyway, and
> finding the bugs involves digging in, as in this host name resolution
> problem...

I'll work on it too, as I'm also testing the stack.

> So, what is the conculsion to the immediate issue of fixing the
> segmentation violations?  Do we want to do a quick fix for that
> (either mine or Waldemars) or leave it as is until the whole of
> libnet.so is cleaned up?  My vote is to put the short term fix in
> now, so I and others can test the rest of the stack without having to
> type in IP addresses :).

If you're coding up test programs, try calling initialize_before() 
before working the stack (?)

> Thanks,
>
> Brennan
> ---- Original Message ----
> From: Waldemar.Kornewald@xxxxxx
> To: openbeosnetteam@xxxxxxxxxxxxx
> Subject: RE: [openbeosnetteam] Re: Fix for resolving host names
> Date: Mon, 2 Feb 2004 17:21:42 +0100
>
> > > That sounds like a big mess to me.
> >
> > The whole stack is one BIG mess, but we do not care anymore. :)
> > It seems that getting a nice stack is the objective for R2. But
> > actually, libnet.so will (or should) be reused, so that one could be
> > cleaned up now. Any volunteers?
> >
> > > 1) what is h_errno actually used for?
> >
> > It is used to report the error status, I think. ;)

:-D

> > > 2) if it's used like the POSIX errno variable, why not copy the
> > > exact mechanism?
> >
> > I will try to find out how it is used exactly (or if it is used at
> > all). Why is there h_errno_tls then?
> > Is our POSIX errno thread-specific or global?

Waldemar's implementation of h_errno is actually correct.

Here's the official definition of h_errno:
  http://www.opengroup.org/onlinepubs/007904975/functions/h_errno.html

What I find interesting in the POSIX definition of h_errno is this 
sentence: "It is unspecified whether h_errno is a macro or an 
identifier declared with external linkage. If a macro definition is 
suppressed in order to access an actual object, or a program defines an 
identifier with the name h_errno, the behavior is undefined."

Nice.

A cursory look in the OBOS sources reveals that we use h_errno_tls for 
thread-specific data.  I'm wondering if you're seeing a segfault 
because h_errno_tls hasn't been instantiated yet (via 
initialize_before() or some other mechanism; this is pure speculation 
on my part.

BTW: The return from hstrerror() looks really weird and isn't 
thread-safe.

Cheers,
Scott