The file resolv/res_init.c in the glibc-2.0.7 tree has
some conditional code which depends on RESOLVSORT being
Building glibc-2.0.7 from the SRPM results in an RPM without
the "sortlist" functionality because the RESOLVSORT isn't
To fix the problem, I just added a -DRESOLVSORT in the
resolv/Makefile and rebuilt the rpms.
It would be great if you could include this define in
future releases of the glibc-2.0.7-X.rpm
That code is experimental and not enough tested. Saying that "it works
for me" is not a guarantee of stability...
------- Email Received From "Mark A. Bentley" <bentlema.edu> 03/23/99 09:15 -------
"Won't fix it" isn't really a useful approach to the problem --
firstly, the missing functionality is documented in resolver(5), so
its omission from libc constitutes at least a documentation bug.
Secondly, the code is not particularly difficult to audit and test.
Finally, the functionality is quite useful.
Could you at least leave this job open as a reminder?
Perhaps you could make alternative rpms that people who use this
feature could "install at their own risk."
I've been using the sortlist feature for about a month now without any
problems on the 61 RedHat machines I maintain at our site.
In any case, if you don't feel that testing this code it RedHat's
responsibility, what would it take to get this included some time
in the future? If the glibc people said that "yes, this code has
been tested, and their have been no reported problems" would this
been good enough for RedHat?
If the glibc people will be happy then they will enable the code by
default. Until then, we are playing it safe.
------- Additional Comments From 09/16/99 19:07 -------
I would agree with bentlema on this one. If you refuse to recognize
the sortlist functionality then at a minimum this should be a doc
bug against the manpage. resolver(5) specifically. Otherwise, I
will assure you that this feature does work. Sun Solaris, and Early
versions of IRIX support it no problem.
*** Bug 5173 has been marked as a duplicate of this bug. ***
See bug 1467, I would have re-opened it if I knew how.
Either fix the resolver(5) manpage, or include support
for the sortlist option. I would prefer the latter but
it's really up to you on that.
Once again, it isn't turned on because this is not the default
behaviour and the extra work required is in most cases not