cvs commit: src/usr.sbin/bsnmpd/modules/snmp_atm Makefile

[Harti Brandt]

On Mon, 23 May 2005, Ruslan Ermilov wrote:

RE>On Mon, May 23, 2005 at 12:24:46PM +0000, Hartmut Brandt wrote:
RE>> harti       2005-05-23 12:24:46 UTC
RE>> 
RE>>   FreeBSD src repository
RE>> 
RE>>   Modified files:
RE>>     usr.sbin/bsnmpd/modules/snmp_atm Makefile 
RE>>   Log:
RE>>   This module was wrongly including header from /usr/include. Give it
RE>>   the correct path to the source headers.
RE>>   
RE>>   Revision  Changes    Path
RE>>   1.2       +1 -1      src/usr.sbin/bsnmpd/modules/snmp_atm/Makefile
RE>> 
RE>It wasn't wrong: during buildworld we first install fresh headers (and
RE>libraries) into a ${WORLDTMP}, then build special versions of compiler
RE>and binutils that pick up things from this location, then build
RE>everything.  You can easily verify this by checking the contents of
RE>the corresponding .depend file (after "make depend" part of "make
RE>buildworld" is done).

Ok, yes. I remember this. I had a report from someone complaining that it 
was including the wrong headers. I think I revert this on the next import 
(I cannot just change the Makefile, but need to change the source too).

RE>It can seem to be wrong when you do standalone compiles though, but
RE>then all you need to do is to install the headers and libraries first.
RE>Think of it: if you have a library that you need to link with, with
RE>the namespace under /usr/lib, your only option when you do standalone
RE>compiles is to rebuild/reinstall this library first.  For headers it
RE>is very much similar.

I see. Thanks for the explanation.

harti