On 09/06/2013 11:25 AM, Konstantin Olchanski wrote:
> Hi, there! Nightly updates of SL6.4 are bombing because of missing boost-devel.i686 update in fastbugs:
>
> --> Finished Dependency Resolution
> Error: Package: boost-devel-1.41.0-17.el6_4.i686 (@triumfcs-mirror-sl-fastbugs)
>             Requires: boost = 1.41.0-17.el6_4
>             Removing: boost-1.41.0-17.el6_4.x86_64 (@triumfcs-mirror-sl-fastbugs)
>                 boost = 1.41.0-17.el6_4
>             Updated By: boost-1.41.0-18.el6.x86_64 (triumfcs-mirror-sl-fastbugs)
>                 boost = 1.41.0-18.el6
>             Available: boost-1.41.0-11.el6_1.2.x86_64 (sl)
>                 boost = 1.41.0-11.el6_1.2
>             Available: boost-1.41.0-15.el6_4.x86_64 (sl-security)
>                 boost = 1.41.0-15.el6_4
>   You could try using --skip-broken to work around the problem
>
> Indeed, the update -devel.i686 is absent:
>
> [root@titan01 ~]# yum list boost-devel boost-devel.i686
> Loaded plugins: refresh-packagekit, security
> Installed Packages
> boost-devel.i686                                1.41.0-17.el6_4                              @triumfcs-mirror-sl-fastbugs
> boost-devel.x86_64                              1.41.0-17.el6_4                              @triumfcs-mirror-sl-fastbugs
> Available Packages
> boost-devel.x86_64                              1.41.0-18.el6                                triumfcs-mirror-sl-fastbugs
> [root@titan01 ~]#
>
>
> P.S. Time to renew my eternal complaint -
>
> ?!?WHY DO BUSTED UPDATES TO SOME JUNK PACKAGE BLOCK ALL SECURITY UPDATES?!?
>
> Why can't "they" fix it already now (in yum, in yum-autoupdate, etc)?
>
> (The CERN version of yum-autoupdate does not block security updates).
>
>
> P.P.S. Yes, right, there was this bike accident a while back, so maybe no fix in yum. RIP.
>
>

The upstream release of boost (BA-2013:1187-1) does not list 
boost-devel-1.41.0-18.el6.i686.rpm within the x86_64 tree.  Thus it was not 
placed in the x86_64 tree.  We have followed the published upstream package 
lists.  When this was published repoclosure did complain which is why I 
verified the package list against upstream before announcing the update.

Is their documentation inaccurate in this regard - does it not match what 
they've actually done?

As for the CERN package, their code is significantly different from the SL 
package.  Their update script appears to utilize --skip-broken on all update 
operations (I've not fully read the script at this time, but this appears true 
for SLC5 and SLC6 systems).  Is this a feature you wish to be available for 
the SL package?  Since I'm not aware of any previous discussion on this and 
have only spent a few moments thinking about it while my lunch microwaves, I 
officially commit to nothing.  If you wish to open a dialog on this, you may.


Pat


PS, unrelated to boost, I've got a few patches posted to fedora's bugzilla for 
adding a few of the yum-autoupdate features to yum-cron-3.4.3.  It already has 
a boolean to enable 'skip-broken' in its configuration files.  I don't believe 
that version can be loaded onto any SL systems - requires python 2.7.

-- 
Pat Riehecky

Scientific Linux developer
http://www.scientificlinux.org/

My mother always said, "When you think someone should do
something, don't forget you are a someone too."