We've talked about special proven testers testing procedures for kernels, and now another candidate for special test procedures has emerged: PackageKit. A bug in PackageKit got out in a Fedora 13 update which broke update notification, which is obviously a big problem. PackageKit isn't actually on the critical path for F13, but it is for F14 and up.
So, I'd like to propose a special testing procedure for updates to PackageKit (actually, any package management-related critpath update should be tested in this way). As well as testing package installation, we should test update notification and installation for such updates. We can consider different ways of doing this, but an easy way would be simply to 'yum downgrade' an installed package, then wait: within a short time (I'll have to check how long with Richard, I think 30 minutes or an hour), an update notification should appear to let you know an update is available to that package. If not, we have problems.
If everyone thinks this makes sense, I'll propose a wiki draft which sets up a system for package-specific testing procedures and drafts such a procedure for PackageKit. What does everyone think of this? Thanks!
+1
Makes perfect sense to me and I think it would be a wise move.
-AdamM
On Thu, 29 Jul 2010 17:01:36 -0400, Adam Williamson awilliam@redhat.com wrote:
As well as testing package installation, we should test update notification and installation for such updates. We can consider different ways of doing this, but an easy way would be simply to 'yum downgrade' an installed package, then wait: within a short time (I'll have to check how long with Richard, I think 30 minutes or an hour), an update notification should appear to let you know an update is available to that package. If not, we have problems.
+1
Sounds good to me, and preventing breakage is a Good Thing (TM)
On Thu, 29 Jul 2010 17:13:50 -0400 "James Cassell" fedoraproject@cyberpear.com wrote:
On Thu, 29 Jul 2010 17:01:36 -0400, Adam Williamson awilliam@redhat.com wrote:
As well as testing package installation, we should test update notification and installation for such updates. We can consider different ways of doing this, but an easy way would be simply to 'yum downgrade' an installed package, then wait: within a short time (I'll have to check how long with Richard, I think 30 minutes or an hour), an update notification should appear to let you know an update is available to that package. If not, we have problems.
+1
Good idea, me thinks.
Subject: Proven tester special testing procedures: PackageKit? From: awilliam@redhat.com To: test@lists.fedoraproject.org Date: Thu, 29 Jul 2010 14:01:36 -0700
We've talked about special proven testers testing procedures for kernels, and now another candidate for special test procedures has emerged: PackageKit. A bug in PackageKit got out in a Fedora 13 update which broke update notification, which is obviously a big problem. PackageKit isn't actually on the critical path for F13, but it is for F14 and up.
So, I'd like to propose a special testing procedure for updates to PackageKit (actually, any package management-related critpath update should be tested in this way). As well as testing package installation, we should test update notification and installation for such updates. We can consider different ways of doing this, but an easy way would be simply to 'yum downgrade' an installed package, then wait: within a short time (I'll have to check how long with Richard, I think 30 minutes or an hour), an update notification should appear to let you know an update is available to that package. If not, we have problems.
If everyone thinks this makes sense, I'll propose a wiki draft which sets up a system for package-specific testing procedures and drafts such a procedure for PackageKit. What does everyone think of this? Thanks! -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org http://www.happyassassin.net
-- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
On 29 July 2010 23:01, Adam Williamson awilliam@redhat.com wrote:
If everyone thinks this makes sense, I'll propose a wiki draft which sets up a system for package-specific testing procedures and drafts such a procedure for PackageKit. What does everyone think of this? Thanks!
Makes sense to me, I certainly need all the help I can get. For the next release and onwards I'm going to try and keep the session updater more in sync with the daemon version (with an artificially high BR) to try and keep the testing matrix of old versions to new versions sane. Thanks guys.
Richard.
On Fri, Jul 30, 2010 at 6:01 AM, Adam Williamson awilliam@redhat.com wrote:
If everyone thinks this makes sense, I'll propose a wiki draft which sets up a system for package-specific testing procedures and drafts such a procedure for PackageKit. What does everyone think of this? Thanks!
+1 Sounds good to me. It's good idea for keeping Fedora better!
On Thu, Jul 29, 2010 at 5:01 PM, Adam Williamson awilliam@redhat.com wrote:
. We can consider different ways of doing this, but an easy way would be
I think that a reasonable way of accomplishing this (I may be insane though :) ) is to have a repo that has a newer version of a package that everyone has installed (say fedora-release) that could be enabled by the tester (perhaps the repo definition could even be in a tester-setup package).
This package could also contain scripts to validate the test setup of the machine in various ways, and to remediate if required. Not sure what a good set of "tests for the tester" is, but food for thought.
On 07/30/2010 02:31 AM, Adam Williamson wrote:
We've talked about special proven testers testing procedures for kernels, and now another candidate for special test procedures has emerged: PackageKit. A bug in PackageKit got out in a Fedora 13 update which broke update notification, which is obviously a big problem.
Would sending a note to fedora-announce with more details and the followup from QA on what is being done to avoid this problem in the future as well.
Rahul
On 07/29/2010 04:01 PM, Adam Williamson wrote:
A bug in PackageKit got out in a Fedora 13 update which broke update notification, which is obviously a big problem. PackageKit isn't actually on the critical path for F13, but it is for F14 and up.
Is there a big tracking this? Was it discussed on IRC only? I see nothing in the PackageKit bodhi update or devel/testing lists mentioning this until your e-mail, which is not very detailed.
I just ran today's updates with gnome-packagekit successfully. On one of my home systems, where I do a majority of my QA time, I saw no issue and reported positive karma. I only use gnome-packagekit on that system specifically to catch any bugs.
Thanks, Michael
On Fri, Jul 30, 2010 at 10:32 PM, Michael Cronenworth mike@cchtml.com wrote:
On 07/29/2010 04:01 PM, Adam Williamson wrote:
A bug in PackageKit got out in a Fedora 13 update which broke update notification, which is obviously a big problem. PackageKit isn't actually on the critical path for F13, but it is for F14 and up.
Is there a big tracking this? Was it discussed on IRC only?
On Thu, 2010-07-29 at 14:01 -0700, Adam Williamson wrote:
We've talked about special proven testers testing procedures for kernels, and now another candidate for special test procedures has emerged: PackageKit. A bug in PackageKit got out in a Fedora 13 update which broke update notification, which is obviously a big problem. PackageKit isn't actually on the critical path for F13, but it is for F14 and up.
So, I'd like to propose a special testing procedure for updates to PackageKit (actually, any package management-related critpath update should be tested in this way). As well as testing package installation, we should test update notification and installation for such updates. We can consider different ways of doing this, but an easy way would be simply to 'yum downgrade' an installed package, then wait: within a short time (I'll have to check how long with Richard, I think 30 minutes or an hour), an update notification should appear to let you know an update is available to that package. If not, we have problems.
If everyone thinks this makes sense, I'll propose a wiki draft which sets up a system for package-specific testing procedures and drafts such a procedure for PackageKit. What does everyone think of this? Thanks!
Quick update on this - I've been ambushed by F14 Alpha stuff. I certainly intend to move this forward when I have time, but if anyone else feels inclined to draft something up, please don't wait for me, go right ahead and do it! Thanks.