Hello Marius, +1 on a KGS 3.4.1 Local overrides here: zope.component = 3.5.1 zope.contenttype = 3.4.2 zope.i18n = 3.5.0 zope.sendmail = 3.5.0 zope.server = 3.5.0 zope.session = 3.5.2 zope.tal = 3.5.0 zope.testing = 3.7.5 #was 3.7.0 but let's try this zope.traversing = 3.5.0a4 zope.testbrowser = 3.5 #mechanize changed the HTTPError class zope.app.appsetup = 3.6.0 zope.app.container = 3.6.0 zope.app.form = 3.5.0 zope.app.intid = 3.5.0 zope.app.publisher = 3.5.0a4 zope.app.wsgi = 3.4.2 #fixes product config zope.app.http = 3.4.5 #fixes IDefaultView vs. PROPFIND lxml = 2.2.4 #or something that works on win32 I might backport a bugfix in zope.publisher r101467 zc.buildout is a tough decision, because it's "not upgrading" issue. See http://mail.python.org/pipermail/distutils-sig/2010-March/015794.html What's the schedule? Wednesday, March 31, 2010, 2:50:33 PM, you wrote: MG> On Wed, Mar 31, 2010 at 01:37:47PM +0200, Christian Theune wrote:
On 03/31/2010 12:35 PM, Adam GROSZER wrote:
One point for the next agenda: As Marius also had the issue to do a KGS 3.4.1.
MG> Mostly I wanted to know if anybody was using the KGS in production and MG> interested in a point release. IRC seemed the right place for a straw MG> poll before an email to zope-dev@, except that I had to run sooner than MG> I thought so I didn't see how people reacted. MG> I recently looked at the zope.release 3.4 branch and saw that it has a MG> bunch of updated package versions since the 3.4.0 release (which was in MG> January 2009). I'm especially interested in setuptools 0.6c11, since MG> the KGS currently pins to 0.6c9, which doesn't support Subversion 1.6 MG> checkouts. MG> We use the 3.4 KGS in production with a few extra pins. Some fix MG> important bugs: MG> zope.app.component = 3.4.2 # very important bugfix for BBB MG> ZODB3 = 3.8.4 # security fixes MG> zope.sendmail = 3.5.1 # This version handles the 5xx errors MG> zope.security = 3.4.2 # bugfixes for Python 2.5 MG> setuptools = 0.6c11 MG> then there are the "we just want the latest tools" pins for zc.buildout MG> and zope.testing, which I'm not proposing to push into a KGS point MG> update (especially with the zope.testing.doctest deprecation which MG> pushes people into using buggy stdlib's doctest). MG> I'm also interested in a bugfix for zope.sendmail that's not done yet: MG> currently it starts the background mail processing thread even if do MG> things like bin/development-instance debug. This can result in emails MG> sent out twice. It also doesn't let the process quit, due to changes in MG> Python's 2.5 threading semantics (atexit handlers that are trying to MG> stop the background thread now are postponed until after all threads MG> terminate). The latest zope.sendmail can avoid this mess by disabling MG> the delivery thread and using a standalone thread-watching script, but MG> it depends on newer core packages and is incompatible with the 3.4 KGS. MG> The 3.4 buildbot tracks the tip of the zope.release's 3.4 branch. All MG> tests pass there. I see that zope.app.component 3.4.2 and setuptools MG> 0.6c11 are already in the (unreleased) 3.4 KGS; ZODB3 3.8.4, MG> zope.sendmail 3.5.1 and zope.security 3.4.2 aren't. MG> So, my plan of action is: MG> * fix zope.sendmail using the approach discussed on this list a few MG> weeks ago MG> * look at other possible bugfix upgrades, see if there are any MG> important bugs fixed (zope.release's bin/list-latest is useful here) MG> * bump ZODB3, zope.sendmail, zope.security versions in the 3.4 KGS, see MG> what buildbot things of this MG> * try to get a 3.4.1 release out of the door <-- this is where I'm MG> fuzzy. I think I used to have ssh access to download.zope.org, but I MG> don't even remember how you're supposed to use zope.release's bin/upload, MG> and I never knew how the releases were made. MG> Marius Gedminas -- Best regards, Adam GROSZER mailto:agroszer@gmail.com -- Quote of the day: Top ten things you'll never hear in an internet chat room: 6. You know, to get a really fast connection, you've gotta go through America Online.