#acl PaulHowarth:read,write,admin,revert,delete All:read

=== Wednesday 15th July 2009 ===

==== Problems with Yum and Arch/Noarch Packages ====

The `yum` tool as shipped in Fedora 11 will choose an arch-specific package over a "later" arch-independent
package - [[RedHatBugzilla:502401|Bug #502401]]. This is a bit of a blow for those of us trying to
move ahead with FedoraProject:Features/NoarchSubpackages and it's been closed `UPSTREAM` and ''not'' fixed in Fedora
(a pet hate of mine). So I built the Rawhide version of `yum` (which contains the fix) for F-11
and stuck it in my local repository. This allowed me to build the `perl-IO-Zlib` update today
(the package has build requirements that have changed from arch-specific to arch-independent,
and F-11's `yum` was selecting the arch-specific version, which was too old for `perl-IO-Zlib`).

Using this update was not without its consequences. My custom `repocheck` script
(a modified version of `repoclosure`) broke:
{{{
Running: repocheck --config=/home/paul/lib/repoc-rhel5-i386.conf --newest --checkrepo city-fan.org-rhel5-i386
Traceback (most recent call last):
  File "/home/paul/bin/repocheck", line 258, in <module>
    main()
  File "/home/paul/bin/repocheck", line 158, in main
    my = RepoClosure(arch = opts.arch, config = opts.config, builddeps = opts.builddeps)
  File "/home/paul/bin/repocheck", line 69, in __init__
    self.doConfigSetup(fn = config,init_plugins=False)
  File "/usr/lib/python2.6/site-packages/yum/__init__.py", line 168, in doConfigSetup
    return self.conf
  File "/usr/lib/python2.6/site-packages/yum/__init__.py", line 681, in <lambda>
    conf = property(fget=lambda self: self._getConfig(),
  File "/usr/lib/python2.6/site-packages/yum/__init__.py", line 210, in _getConfig
    arch = self.arch.canonarch
AttributeError: 'list' object has no attribute 'canonarch'
}}}
Now I don't know much `python` but after a bit of googling around and looking at the code,
I made these changes, which seemed to fix things:
{{{
--- repoclosure.orig	2009-07-15 17:21:13.168971727 +0100
+++ repoclosure	2009-07-15 17:26:10.748969996 +0100
@@ -69,10 +69,11 @@
     def __init__(self, arch=[], config="/etc/yum.conf", builddeps=False, pkgonly=None):
         yum.YumBase.__init__(self)
         self.logger = logging.getLogger("yum.verbose.repoclosure")
-        self.arch = arch
+        self.arch.archlist = arch
         self.builddeps = builddeps
         self.pkgonly = pkgonly
-        self.doConfigSetup(fn = config,init_plugins=False)
+        self.preconf.fn = config
+        self.preconf.init_plugins = False
         if hasattr(self.repos, 'sqlite'):
             self.repos.sqlite = False
             self.repos._selectSackType()
@@ -98,10 +99,10 @@
     def readMetadata(self):
         self.doRepoSetup()
         archs = []
-        if not self.arch:
+        if not self.arch.archlist:
             archs.extend(rpmUtils.arch.getArchList())
         else:
-            for arch in self.arch:
+            for arch in self.arch.archlist:
                 archs.extend(rpmUtils.arch.getArchList(arch))
         if self.builddeps and 'src' not in archs:
             archs.append('src')
}}}

==== Fedora Project ====

 * Updated `gtkwave` for F-11 and Rawhide to resolve [[RedHatBugzilla:511858|Bug #511858]], an easily-reproducible crash when trying to use ''Print to File''
 * Looking through [[http://www.redhat.com/mailman/listinfo/fedora-perl-devel-list|fedora-perl-devel-list]], I noticed a bunch of FTBFS (Fails To Build Build From Source) bugs had been raised. I didn't have time to look through many of them, but I noticed that the `perl-Template-Plugin-Class` failure had also been reported upstream and that a new version 0.14 had been issued, removing the bogus test that had triggered the failure. So I updated `perl-Template-Plugin-Class` and closed [[RedHatBugzilla:511438|Bug #511438]] accordingly.

==== Local Packages ====

 * Updated `gtkwave` as per the Fedora version
 * Updated `perl-IO-Zlib` to 1.10 (no functional changes)

----