[ntp:hackers] Odd warning in ntp_rfc2553.c

mayer mayer at gis.net
Sun May 22 12:00:55 PDT 2005


----- Original Message Follows -----
> I am working on integrating NTP V4 into Solaris, which at this
> stage mostly means morphing the directory structure and Makefiles
> to conform to our internal build requirements.
> 
> Anyway, I get this warning when compiling ntp_rfc2553.c:
> 
> ntp_rfc2553.c", line 287: warning: const object should have
> initializer:  isc_net_in6addrany (E_CONST_OBJ_SHOULD_HAVE_INITIZR)
> 
> This seems to be due to this section of code:
> 
> #if defined(SYS_WINNT)
> /* XXX This is the preferred way, but for some reason the SunOS
> compiler
>   * does not like it.
>   */
> const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
> #else
> const struct in6_addr in6addr_any;
> #endif
> 
> 
> Now, granted, we set the compilers to issue warnings about a lot
> of things, and I don't get this error if I turn off all the
> checking, but I am intrigued about how to resolve this. I have tried
> the oldest Sun Compiler I can find, and they all ignore this
> problem without the error checking and flag it with the error
> checking. No version I can find complains the other way.
> 
> So, I have to assume that the comment is from older SunOS 4.x
> compilers. Now, my problem is two fold. It is my desire to
> make the compile run clean, but also to reduce my down stream
> maintenance. To that end, any changes I need to make, I would like
> to see also made in the ntp-dev code base, so that I do not have
> to go through this whole thing at every code release. But, I
> don't see any obvious flags that I can check for to add to the
> ifdef that will use the first version on Solaris and the second
> on SunOS. In fact, it seems odd to me that what the comment
> calls the "preferred way" is only used on WINNT. Shouldn't
> the sense of the ifdef be the other way around so that
> only SunOS uses the line without the initializer? There are
> comments in the distribution about the desire to remove
> the obvious choice, SYS_SOLARIS, and as near as I can tell,
> this symbol isn't set anyway. So, any ideas about what to do
> and why this ifdef is there in the first place?
> 
> What is the status of support of SunOS anyway? Could we just get
> rid of the whole ifdef?
> 

Brian, can you file a bug report on this. This is one of those issues
that never quite gets resolved. I suspect that recent versions of
Solaris don't really need this initialization as it's somewhere else.
However I don't know that so you will need to tell us.

Danny

> -- 
> blu




More information about the hackers mailing list