[Bug 226192] Merge Review: net-snmp

[bugzilla at redhat.com]

Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=226192

--- Comment #8 from Jan Safranek <jsafrane at redhat.com> 2010-01-12 07:33:28 EST ---
(In reply to comment #7)
>   net-snmp.x86_64: W: dangerous-command-in-%post mv
> Agreed that this may be OK, but what was the last release that put data in
> /var/net-snmp?  If it was more than a few releases ago, there's no point in
> keeping this bit at all.  At the least, some indication of which releases
> require it is in order.

Fedora 11: /var/net-snmp, Fedora 12: /var/lib/net-snmp. I've added a note about
it.


> Similarly, "allow compilation on old Fedoras" can go if "old Fedoras" is older
> than F10.  At the very least, some statement of "how old" would be good to
> have.

Again, rpm on Fedora 11 does not define %python_sitearch, I've added a comment
there.


>   net-snmp-libs.x86_64: W: undefined-non-weak-symbol
>    /usr/lib64/libnetsnmpagent.so.20.0.0 netsnmp_register_null_context
> Six of these in this library with different symbols.  I suppose if you want to
> use this library you also have to link against some other library which
> provides those symbols.  That's bad programming practice but not a review
> blocker.

This one is broken upstream, libnetsnmpagent.so does not link with
libsnmphelpers.so, but needs its symbols. In fact, libsnmphelpers and
libnetsnmpagent need symbols from each other :). I've filled a bug upstream and
I'll look at it eventually.
https://sourceforge.net/tracker/?func=detail&atid=112694&aid=2930536&group_id=12694


>   net-snmp-libs.x86_64: W: unused-direct-shlib-dependency
>    /usr/lib64/libnetsnmpagent.so.20.0.0 /usr/lib64/libcrypto.so.10  
> Many of these in various different libraries.  All of these libraries are
> linked against other libraries even though they call nothing in them.  This
> probably isn't an issue as most of these libraries will be in memory anyway,
> although there are a couple of libperl.so references that don't need to be
> there.
> In general these are OK; they're bad if the force unnecessary dependencies or
> pull in unused libraries that won't generally already be in memory, and you can
> get rid of them with "-Wl,--as-needed" on the link line.  Maybe that's what
> "--enable-as-needed" on the configure line is for, but it doesn't seem to help.

Strange, it looks like --enable-as-needed does absolutely nothing. I'll check
it with upstream, this looks like a silly bug.


>  In any case I don't see anything which would block a review.
Thanks a lot for the review! Would you kindly approve it?


> One minor nit is that I don't see the point of the "Building option" bit in
> %description.  When you see that in the final package, it makes it seem that
> use of tcp_wrappers is disabled when in fact it's enabled.  Such a comment is
> appropriate for the spec file but not really useful in the binary package.

"Building option" removed.

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.