On Feb 18, 2012, at 1:55 PM, Daniel Macks wrote: > On Sat, 18 Feb 2012 13:01:21 -0500, Charles Lepple <clep...@gmail.com> wrote: > On Feb 17, 2012, at 9:22 PM, Alexander Hansen <alexanderk.han...@gmail.com> > wrote: >> >> > Check if you've got net-snmp15-dev or net-snmp25-dev installed, and if >> > so remove it. >> I vaguely recall some resistance to using BuildConflicts to force using the >> system net-snmp, but there is no reason why it couldn't use the Fink SNMP >> libraries instead. Any preference? > > There is a bug in the net-snmp* package-suites, that /sw/lib/libnetsnmp.la > (which is what libtool is finding when nut uses -lnetsnmp) is in net-snmp > rather than the -dev package (the correct place and the one one would expect > to BCon to avoid a lib) that contains the actual .dylib. I'm about to commit > a fixed set of net-snmp* packages. So you'd have to BCon on several -dev > (there are multiple libversions and also package renamings in the supported > distros) and also on the main %N for each of them. But only the older %v-%r > of the latest of %N because %N also contains user-land programs and other > runtime stuff that users would be sad and confused if they got uninstalled. > BCon is always a crappy crutch of a solution precisely because it breaks > every time the actual thing you're trying to avoid changes into a different > package. If you want to avoid a lib, need to patch the actual configure > and/or makefiles and/or ...whatever else the package uses. Or else maybe > easier to make it actually use the fink lib.
So... would it be better for NUT to just depend on a single Fink net-snmp package (>= fixed rev) since that is more likely to conflict with other versions of net-snmp? I think my original motivation for linking against the system version was just to avoid compiling all of net-snmp if it wasn't going to be used. Processors have gotten faster since then, but given that the original bug report was for PowerPC/10.5, maybe it makes sense to split the SNMP driver off completely from this info file (so that it doesn't get compiled at all for non-SNMP UPSes). Or, if we're doing something wrong in the NUT configure script to detect net-snmp, let me know and I'll try to get it fixed upstream. I don't use the SNMP support myself, but I can probably get someone to test changes. >> I'll look at that linker error you mentioned in the other thread, too. Can't >> reproduce the error, but I think it's wrong to declare variables (without >> extern) in a header file. > > That's exactly what it's doing (about a half-dozen of them I found). There > were also some stray deps no-longer-needed by recent revisions of the gd2 > package. I'll apply the extern fixes upstream as well. -- Charles Lepple clepple@gmail ------------------------------------------------------------------------------ Virtualization & Cloud Management Using Capacity Planning Cloud computing makes use of virtualization - but cloud computing also focuses on allowing computing to be delivered as a service. http://www.accelacomm.com/jaw/sfnl/114/51521223/ _______________________________________________ Fink-users mailing list Fink-users@lists.sourceforge.net List archive: http://news.gmane.org/gmane.os.macosx.fink.user Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-users
