The package currently has three bugs open: #249016 (big file in filesdir – a whole SVN revision), #248409 (--as-needed failure) but most importantly #228187 (fails to build with current stable glibc – opened over an year ago). In addition to this, there are a few other problems that makes it a candidate to be removed ASAP: - the whole filesdir is big, it's the second biggest after glibc!; - mistakes the use of debug USE flag, fiddling with optimisation and symbol flags; - parallel make is disabled, no bug open for reporting that problem; - make does not die in _compile!; - cross-filesystem symlinks for /sbin!; - installs basically _all_ the tests as part of libexec (uh?); - installs libraries in /usr/lib/$PN and adds that to the ldpath.. (uh?!) Now, this was any other package I would have sent the QA last rites already, unfortunately the reverse deps tree of openais are quite deep, which means I cannot simply merry-go-round remove this from the tree. sys-cluster/cman and sys-cluster/fence depend on this, and those are dependencies of another huge set of ha-cluster packages. I'm leaving 15 days to ha-cluster team to decide whether to pick up this package from Wolfram (who don't seem to be interested in keeping it up to speed afaics). If this bug does not at least provide a feasible course of action on July 6th, then I'm going to proceed with the _recursive_ masking of openais, and recursive removal in 60 or 90 days (I might decide to extend the grace period given the amount of stuff to verify).
Is anyone alive on ha-cluster? It seems like we should get the word out that we need more people for this team?
Seems like we're going to start with the masking.
clvm is obsolete anyway, it's functionality has been merged with the main lvm2 package. rocket just never ever responded to me about officially migrating the users. This removal does however leave us without any clustered disk backend for LVM2. Of the three options: gulm, cman, corosync - GULM was removed by ha-cluster - and corosync was never in the tree
Well, technically we haven't had cman for a long time given openais does not compile..
The cman use flag needs to be masked as well to avoid breaking the deps: sys-fs/lvm2-2.02.48-r1: nonsolvable depset(depends) keyword(~x86) profile (default/linux/x86/2008.0/desktop): solutions: [ =sys-cluster/cman-2* ] for example.
I'm quite sure I did that too.
Forgot to remove the unmasking in arch/x86/use.mask I fixed it for you.
Ah thanks, I didn't expect it to be unmasked anywhere, sorry.
I'm waiting for the new stable openais/corosync packages to be released (1.0.0) soon... I don't use the old stuff anymore.
FYI: (Jul 8, 2009) OpenAIS 1.0.0 stable released!!!
I can probably maintain openais and much of gfs2 in funtoo, but I'm not a developer. I will need someone to pull my changes from funtoo into gentoo.
Quick fix for openais-0.80.3-r1.ebuild is to add an "unset LDFLAGS" and "make" to src_compile(). parallel make works fine. I have not tested was was originally in the ebuild's src_compile(). Note that the LDFLAGS problem seems to be introduced by files/Makefile.inc-FLAGS.patch.
I quickly hacked something in bug #279631 I hope that the ha team revives sooner or later ^^
wschlich: you were waiting for 1.0.0, please get it into the tree now, see the gentoo-cluster mailing list of people wanting it.
(In reply to comment #14) > wschlich: you were waiting for 1.0.0, please get it into the tree now, will try to accomplish that until 2009-12-06. > see the gentoo-cluster mailing list of people wanting it. http://archives.gentoo.org/gentoo-cluster/ -- nothing?! :)
Here's the guy that needed it: http://archives.gentoo.org/gentoo-cluster/msg_deef03881d2b0bb9f47ca3120eb8e87e.xml
removed from main tree