Klemens Nanni <[email protected]> wrote:

> On Wed, Jul 10, 2019 at 03:44:55PM +0100, Ricardo Mestre wrote:
> > Index: dhcpd.h
> > ===================================================================
> > RCS file: /cvs/src/sbin/dhclient/dhcpd.h,v
> > retrieving revision 1.278
> > diff -u -p -u -r1.278 dhcpd.h
> > --- dhcpd.h 22 May 2019 12:56:31 -0000      1.278
> > +++ dhcpd.h 10 Jul 2019 14:36:31 -0000
> > @@ -153,6 +153,8 @@ struct interface_info {
> >  
> >  #define    _PATH_DHCLIENT_CONF     "/etc/dhclient.conf"
> >  #define    _PATH_LEASE_DB          "/var/db/dhclient.leases"
> > +#define    _PATH_RESOLV_CONF       "/etc/resolv.conf"
> /usr/include/resolv.h:#define _PATH_RESCONF        "/etc/resolv.conf"
> 
> > +#define    _PATH_RESOLV_CONF_TAIL  "/etc/resolv.conf.tail"
> That one is not defined elsewhere.
> 
> We've got eight occurences of _PATH_RESCONF in /usr/src/!(sys), whereas
> "/etc/resolv.conf" used as literal string 33 times throughout base.

resolv.h exports a very large number of symbols.  Some of them have
very short names (MAXNS) and are likely to collide.  A bunch of them
are very domain-specific (like the pun), but also very likely to collide
in ugly ways.

No wonder poeple hesitated to #include this ugly file.

Look, you are bringing up is a different battle.  For the proposed diff,
it is a distraction.  You're not proposing holding up mestre's diff are
you, cause it looks like it..

Anyways, if this was looked into later as an independent step, we could
study whether _PATH_RESCONF was placed in the wrong file and should have
been in <paths.h>, try a move, and see what the fallout looks like.

Reply via email to