Amanda 3.1.0 was released last week - it would be great to have it available in gentoo testing.. Reproducible: Always Steps to Reproduce:
*** Bug 322777 has been marked as a duplicate of this bug. ***
Created attachment 237115 [details] new release, new ebuild for 3.1.1
Looks like 3.1.0 got missed in portage :)
Comment on attachment 237115 [details] new release, new ebuild for 3.1.1 Looks like it was based on something older than the 2.6.1_p2ebuild. Please attach a diff next time or nothing if there is no difference and you just copied over an old ebuild.
FYI, 3.1.2 should be out at the end of this week, so 3.1.1 (or at least 3.1.0) should get into testing soon.
Created attachment 241363 [details] config.log with --with-ipv6 --- amanda-2.6.1_p2.ebuild 2010-05-08 21:48:53.000000000 +0200 +++ amanda-3.1.1.ebuild 2010-08-04 03:55:57.668039175 +0200 @@ -217,10 +217,6 @@ myconf="${myconf} --with-bsdudp-security" myconf="${myconf} --with-bsdtcp-security" - # kerberos-security mechanism version 4 - # always disable, per bug #173354 - myconf="${myconf} --without-krb4-security" - # kerberos-security mechanism version 5 myconf="${myconf} `use_with kerberos krb5-security`" The above gets rid of a minor issue. The bigger issue that keeps me from committing this ebuild is that with USE=ipv6, the configure script actually expects to be able to bind to an IPv6 socket, instead of checking that compile time support is present as it should. config/amanda/ipv6.m4 even spits out "Target system has functional IPv6 support", while we haven't even told configure what the target system is.
The term "target" there should be "host" - Amanda definitely does not support cross-compiling! I'll fix that up (in trunk only, so you won't see the change in 3.1.2). We test this way because it's the only way to effectively determine whether a system can really support an IPv6 installation. Without this check, Amanda can build on systems (like yours) with compiler support for IPv6 but without kernel support. Because many such systems have ::1 in /etc/hosts, Amanda ends up trying to use an IPv6 socket even for inter-process communication, and things fail, badly. There are obviously better fixes - in particular, Amanda could retry failures with AF_INET6 again with AF_INET. In many places, it does so. The problem is that "fixing" these sorts of problems has consistently introduced bugs for IPv4 users (arguably 100% of Amanda's users) and only benefit the IPv6 users (0%). So we've stuck with a conservative approach of just not building IPv6 support if it looks like it might not work at runtime. I wrote about related issues recently: http://code.v.igoro.us/archives/61-IPv6-and-Amanda.html I'd be more than happy to upstream any patches you provide! To the point at hand, though: 1. this exact same code is present in 2.6.1, so it's really orthogonal to bumping 3.1.1, and probably deserves its own bug 2. if your system has compile-time support for IPv6, but no kernel support, then it really doesn't have IPv6 support, right? So why are you passing USE=ipv6? I suppose a good fix is to skip the bind-to-a-socket test if --with-ipv6 was supplied, and just assume the user knows what they're doing. What do you think?
Fix "Target": http://github.com/zmanda/amanda/commit/2126a109369204747ea4bc65ef5bfce8624b0e60
Bumping the version-number target. 3.1.2 was released last week.
Created attachment 243085 [details] ebuild for amanda-3.1.2
Stefan: did you upload the wrong file, by chance? That's byte-for-byte identical to attachment 237115 [details]. Or did you not make any changes? In comment #4, Jeroen pointed out that this is based on an older ebuild than the 2.6.0_p2-r4 that's currently in the tree. In particular, the tree has copyright 1999-2009, while your patch has copyright 1999-2008. There are significant differences between the ebuilds, and I suspect most of them were not your changes, but are regressions. Can you upload a newer ebuild based on the most recent in portage?
Created attachment 243105 [details] corrected ebuild for 3.1.2 shame on me ... new ebuild now, based on current ebuild for 2.6.1p2 in portage, and regarding the corrections by Dustin (from the thread on amanda-users).
(In reply to comment #12) > Created an attachment (id=243105) [details] > corrected ebuild for 3.1.2 > > shame on me ... new ebuild now, based on current ebuild for 2.6.1p2 in portage, > and regarding the corrections by Dustin (from the thread on amanda-users). Reminder to maybe speed up getting this into portage ...... hello? thanks ...
3.1.2 in tree.
