måndag den 11 april 2011 klockan 14:55 skrev Jamie Wilkinson detta: > I think the right solution is to make it a constrained depends on ~kFreeBSD > systems, because I'm pretty sure it doesn't work without iproute on Linux. > (My memory is weak in this area though.)
I have made some quick checks in order to arrive at a sound decision. There is a configuration switch "--enable-ulog" which activates the only code that is dependent on Linux-only headers. It is not enabled in the rules file, so a test build for the architecture i386 produces `uacctd` which only is able to print a trivial message, yet consumes 466 kB! The source file `src/uacctd.c` uses a conditional clause which excludes all interesting code, so the full weight of the produced `uacctd` is linking code! Taking all configuration switches from the rules file, only adding `--enable-threads`, to kfreebsd-amd64 and pmacct-0.14.0rc1 , allows the full binary build to complete successfully. (I have yet to test all executables, but `uacctd` is even bigger and is equally trivial in capability!). When inserting `--enable-ulog` the build step will abort with the expected error when compiling `uacctd`. Thus, the question remains: Is there really something in the official Debian binary package that provably depends on `iproute`. All the more, since `uacctd`, which should depend on the presence of Netfilter, is not even built correctly. Is there some part of the infrastructure of the binary package tailored to Netfilter/iptables/ip6tables, thus conditioning on `iproute`? The bulk of the package consists of compiled binaries, and of SQL schema for the different backends. In passing, let me also wonder if package dependency on `libmysqlclient15-dev` is necessary. Would not `libmysqlclient-dev` suffice? Best regards, Mats E Andersson, DM -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
