Re: bodhi statistics
On Thu, Jun 10, 2010 at 04:59:30AM +0200, Kevin Kofler wrote: Luke Macken wrote: Neither of you have mentioned your definition of the word success. Care to enlighten us? Success is the achievement of a worthwhile goal. If the original goal which was set is worthless, succeeding at it is meaningless. There are a number of people who disagree with the value of that goal, including most of the past and current FESCo. It is very clear you don't agree with the goal. That in and of itself does not make it worthless. Luke's changes to bodhi are also not meaningless. He is implementing towards the goal/process he was given and he's doing it rather well. You are devaluing the work of a Fedora contributor and being generally poisonous. I suggest you stop this particular line of attack. josh -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
Josh Boyer wrote: There are a number of people who disagree with the value of that goal, including most of the past and current FESCo. You conveniently ignored the part of the mail where I pointed out WHY that goal is broken. Why are we trying to emulate the failed Fedora Legacy process rather than the successful Fedora Extras one? Why can't we learn from our past? Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Thu, Jun 10, 2010 at 05:10:36PM +0200, Kevin Kofler wrote: Josh Boyer wrote: There are a number of people who disagree with the value of that goal, including most of the past and current FESCo. You conveniently ignored the part of the mail where I pointed out WHY that goal is broken. No, that was intentional. I was commenting directly on your abuse of someone that was _told_ to implement towards a specific goal. Your questioning of that goal, while personally annoying to me, is certainly acceptable. Your assertion that someone's hard work and time on an implementation of that goal is worthless and meaningless is not. It's hurtful, poisonous and entirely unwelcome. josh -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
* Kevin Kofler [10/06/2010 17:21] : Why are we trying to emulate the failed Fedora Legacy process rather than the successful Fedora Extras one? Why can't we learn from our past? I think you're rather quick to conclude that Extras succeded because it allowed maintainers to push to stable directly and that Legacy failed for the lack of same. Others have given other reasons for Legacy's failure (these have ranged to lack of infrastructure to lack of community involvement to conspiracy theories involving Red Hat), so if stringent QA requirements really were the cause for Legacy's failure, it's probably best to demonstrate this before using it as an arguement in your assertions. Emmanuel -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Thu, 10 Jun 2010 04:59:30 +0200 Kevin Kofler kevin.kof...@chello.at wrote: Now, if the policies that are being approved do not actually benefit the greater good of the community, we have bigger problems. Yet that's exactly the problem we're having. :-( snip... So why are we now going to use the Fedora Legacy model over the Fedora Extras one? Does FESCo really want Fedora to fail? I think you are presenting a false dichotomy here. There are many reasons why the current fedora is dissimilar to both Fedora extras and Fedora Legacy. I think it's worth trying to give our users a more stable updates stream. If it turns out that we don't get this from the things we are trying to do, we will learn from it and try something else. I think we have had this discussion before, so I will leave it at that. kevin signature.asc Description: PGP signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, 2010-06-08 at 16:51 -0400, Luke Macken wrote: Fedora 13 * 231 updates automatically pushed due to karma (6.49%) * 2 of which were critical path updates So I thought that this last metric would take into account critical path updates that were approved under the new policy, but it did not. The actual metric is for Fedora 13 is: * 333 critical path updates approved -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, 2010-06-08 at 21:20 -0400, Matthias Clasen wrote: On Tue, 2010-06-08 at 16:51 -0400, Luke Macken wrote: Fedora 13 * 3562 updates * 3065 stable updates * 427 testing updates * 62 pending updates * 8 obsolete updates Hey Luke, are these the numbers of F13 updates since we branched, or since the F13 release ? The numbers certainly look like the former, but I think the latter would be much more interesting. These numbers are since we branched for F13. luke -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
Luke Macken wrote: By success I mean that I felt we were successful in drafting, implementing, deploying, and utilizing the mentioned policies as expected, and the results show increased community engagement. This definition of success does not match mine nor the one you'll find in a dictionary. So your terminology is misleading. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, 2010-06-09 at 08:38 +0200, Kevin Kofler wrote: Luke Macken wrote: By success I mean that I felt we were successful in drafting, implementing, deploying, and utilizing the mentioned policies as expected, and the results show increased community engagement. This definition of success does not match mine nor the one you'll find in a dictionary. So your terminology is misleading. Really, Kevin? We're digressing to a dictionary battle? Fine, I'll play. First definition in the dictionary[0]: an event that accomplishes its intended purpose. ...which is exactly what I meant to being with. luke [0]: http://www.google.com/search?q=define%3Asuccess -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On 06/09/2010 08:54 AM, Luke Macken wrote: On Wed, 2010-06-09 at 08:38 +0200, Kevin Kofler wrote: Luke Macken wrote: By success I mean that I felt we were successful in drafting, implementing, deploying, and utilizing the mentioned policies as expected, and the results show increased community engagement. This definition of success does not match mine nor the one you'll find in a dictionary. So your terminology is misleading. Really, Kevin? We're digressing to a dictionary battle? Fine, I'll play. First definition in the dictionary[0]: an event that accomplishes its intended purpose. Exactly. Your definition differs from Kevin's (and mine). ...which is exactly what I meant to being with. To me, your definition of success is compliance with *your* process. Whether this process is suitable to improve package quality, whether the technical system behind it is a good approach and whether your approach actually improves package quality or is mere bureaucray is highly questionable. That said, all you demonstrated is your system not being entirely broken, but I don't see any success related to QA in your statistic. Ralf -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On 06/08/2010 10:51 PM, Luke Macken wrote: I recently wrote some code to generate detailed statistics of Fedora EPEL updates within bodhi. Eventually this will be auto-generated and exposed within bodhi itself, but for now here are the initial metrics. This report definitely conveys the shortcomings in our testing, however, it does show us improving with each release. For Fedora 13, we implemented the No Frozen Rawhide process with improved Critical Path policies, which were definitely a success. With these enhanced procedures, along with the upcoming implementation of AutoQA and the new Package update acceptance criteria (https://fedoraproject.org/wiki/Package_update_acceptance_criteria), I think we'll see these numbers drastically improve in the future. I can't agree that this update policy is success (in any dictionary). Since I'm forced to wait two weeks for pushing into stable, I have more tickets about packages that I've already fixed. Users want fixes immediately, they are not interested in some processes. Many users don't even have FAS account and they want fixes and updates now. Not in next release as offer enterprise solution. Other thing is that I'm forgetting packages in bodhi, because I believed until yesterday that updates will be pushed after two weeks automatically. Not sure whether policy changed or it's a bug [1]. Karma in bodhi updates doesn't speak about quality of updates at all. At least my packages which has more dependencies are pushed without testing, and less important are tested even testers don't know what does it do [2]. For example here were testers happy because it was updated without any warning message, but they don't know if it's working or not. I'm looking forward to qa project that could really improve quality - like testing updates, which fixed bugs with bugzilla and testers really reproduce them before and not after. Otherwise is bodhi good only for checking update by yum without errors. Best regards, Marcela [1] https://admin.fedoraproject.org/updates/perl-LDAP-0.40-1.fc13 [2] https://admin.fedoraproject.org/updates/perl-version-0.82-1.fc13 -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, 8 Jun 2010 16:51:36 -0400 (EDT), Luke wrote: = Bodhi Statistics Report (Generated on June 8th, 2010) = Out of 17412 total updates, 2958 received feedback (16.99%) Out of 1045 total unique karma submitters, the top 30 are: Just for the record and some fairness, the top Fedora 11 testers according to https://admin.fedoraproject.org/updates/metrics/?release=F11 alexlan mschwendt mrunge dsd dcottle lmacken mooninite cwickert from times when the fedora-easy-karma script wasn't available and didn't make it possible to mass-vote on updates. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, Jun 08, 2010 at 04:51:36PM -0400, Luke Macken wrote: You can find the code that generates these statistics here: https://fedorahosted.org/bodhi/browser/bodhi/tools/metrics.py https://fedorahosted.org/bodhi/browser/bodhi/tools/log_stats.py. If you have any ideas or suggestions for different types of metrics to generate, or if you find any bugs in my code, please let me know. * 595 unique authenticated karma submitters I think it might be interesting to calculate this number for different timeframes, e.g. last day, last week, last month, since it was released, since F(N+1) was released. Maybe also the number of karma submitters that submitted at least X comments for each week the release is available. This would imho allow more precise to estimate how many active testers are there. Regards Till pgp1oy9tUDVfk.pgp Description: PGP signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, Jun 09, 2010 at 10:58:15AM +0530, Rahul Sundaram wrote: On 06/09/2010 10:48 AM, Kevin Kofler wrote: Stephen John Smoogen wrote: Well the only person I see mentioning quality is Kevin. And for some reason he is expecting it immediately You can't claim that there are shortcomings in our testing (the exact words Luke used!) without a metric of quality. A shortcoming in testing means less testing was done than needed. If the packages have good quality, that means more testing is NOT needed, no matter what the actual amount of testing was. Packages without any testing is of unknown quality. If someone doesn't test them before pushing, users are unknowingly playing that role. That is true if you are making the assumption that the package maintainer did no testing themselves. I would hope that isn't the common case. josh -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, Jun 09, 2010 at 09:51:59AM +0200, Michael Schwendt wrote: On Tue, 8 Jun 2010 16:51:36 -0400 (EDT), Luke wrote: = Bodhi Statistics Report (Generated on June 8th, 2010) = Out of 17412 total updates, 2958 received feedback (16.99%) Out of 1045 total unique karma submitters, the top 30 are: Just for the record and some fairness, the top Fedora 11 testers according to https://admin.fedoraproject.org/updates/metrics/?release=F11 alexlan mschwendt mrunge dsd dcottle lmacken mooninite cwickert from times when the fedora-easy-karma script wasn't available and didn't make it possible to mass-vote on updates. Thanks for this. I was wondering about exactly that. josh -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On 06/09/2010 05:12 PM, Josh Boyer wrote: That is true if you are making the assumption that the package maintainer did no testing themselves. I would hope that isn't the common case. There are well known cases of that happening. Kevin has been public about his position on that. Hopefully we can automatically catch and prevent the obvious breakages soon. Rahul -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wednesday 09 June 2010, Rahul Sundaram wrote: There are well known cases of that happening. Kevin has been public about his position on that. Hopefully we can automatically catch and prevent the obvious breakages soon. For the record, I did test (on one release) the latest bunch of updates I pushed before pushing them out to stable. But sometimes a fix is very urgent and trivial so I'll just take anyone's word that it works (or if the change is really really trivial, e.g. a fixed Requires which can be verified to be satisfied just by looking at the Rawhide report and noticing there's no broken dependency, not even that) and issue it directly to stable (and I think that's a good thing and the new process which will make this a PITA is going to degrade our overall package quality by delaying urgent fixes). And for stuff like the KDE updates, I know many people, including other KDE SIG members, test them, so I usually don't spend my time testing those updates as well, the testing feedback we get is largely sufficient. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
Marcela Mašláňová wrote: I can't agree that this update policy is success (in any dictionary). Since I'm forced to wait two weeks for pushing into stable, I have more tickets about packages that I've already fixed. Users want fixes immediately, they are not interested in some processes. Many users don't even have FAS account and they want fixes and updates now. Not in next release as offer enterprise solution. Exactly my point. But FESCo just won't listen. :-( Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, 9 Jun 2010, Kevin Kofler wrote: Marcela Mašláňová wrote: I can't agree that this update policy is success (in any dictionary). Since I'm forced to wait two weeks for pushing into stable, I have more tickets about packages that I've already fixed. Users want fixes immediately, they are not interested in some processes. Many users don't even have FAS account and they want fixes and updates now. Not in next release as offer enterprise solution. Exactly my point. But FESCo just won't listen. :-( FESCo listened. They did not agree with you. -sv -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, 2010-06-09 at 09:35 +0200, Marcela Mašláňová wrote: On 06/08/2010 10:51 PM, Luke Macken wrote: I recently wrote some code to generate detailed statistics of Fedora EPEL updates within bodhi. Eventually this will be auto-generated and exposed within bodhi itself, but for now here are the initial metrics. This report definitely conveys the shortcomings in our testing, however, it does show us improving with each release. For Fedora 13, we implemented the No Frozen Rawhide process with improved Critical Path policies, which were definitely a success. With these enhanced procedures, along with the upcoming implementation of AutoQA and the new Package update acceptance criteria (https://fedoraproject.org/wiki/Package_update_acceptance_criteria), I think we'll see these numbers drastically improve in the future. I can't agree that this update policy is success (in any dictionary). If you can't agree that we, as a community, were successful in accomplishing our intended purpose of implementing the No Frozen Rawhide Critical Path package processes that we drafted, then you're not understanding the use of the word. Again, the update policy that is currently rubber stamped does not match what is enforced by the process. The package update acceptance criteria[0] has been approved, but not yet implemented. So, if you feel like you have suggestions or some sort of constructive criticism to offer, I recommend bringing it up with FESCo. [0]: https://fedoraproject.org/wiki/Package_update_acceptance_criteria Since I'm forced to wait two weeks for pushing into stable For EPEL updates, yes, that is their policy. For Fedora, there is nothing stopping you from pushing your non-critpath updates straight to stable. Other thing is that I'm forgetting packages in bodhi, because I believed until yesterday that updates will be pushed after two weeks automatically. Not sure whether policy changed or it's a bug [1]. Bodhi has never done this, ever. According to the new acceptance critera, updates will have to spend some minimum amount of time in updates-testing, currently one week. Now, as to whether or not bodhi should auto-push after that week, that I'm not quite sure. Karma in bodhi updates doesn't speak about quality of updates at all. At least my packages which has more dependencies are pushed without testing, and less important are tested even testers don't know what does it do [2]. For example here were testers happy because it was updated without any warning message, but they don't know if it's working or not. Right, karma does not equate quality. I don't think anyone ever claimed that it did? I'm looking forward to qa project that could really improve quality - like testing updates, which fixed bugs with bugzilla and testers really reproduce them before and not after. Otherwise is bodhi good only for checking update by yum without errors. Yes, I'm looking forward to AutoQA as well. We all are. The majority of karma is this ran without exploding, which is far from sufficient testing, but it's still valuable information. luke -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
Luke's dictionary is more correct than yours. anyone else see how horrid the line I just wrote sounded in your head when you read it? That's what this thread sounds like. Did we really need to take some raw numbers that Luke was kind enough to put together and make it into some sort of QA methods holy war? -AdamM (From Android) On Jun 9, 2010 2:11 AM, Ralf Corsepius rc040...@freenet.de wrote: On 06/09/2010 08:54 AM, Luke Macken wrote: On Wed, 2010-06-09 at 08:38 +0200, Kevin Kofler wrote: ... Exactly. Your definition differs from Kevin's (and mine). ...which is exactly what I meant to being with. To me, your definition of success is compliance with *your* process. Whether this process is suitable to improve package quality, whether the technical system behind it is a good approach and whether your approach actually improves package quality or is mere bureaucray is highly questionable. That said, all you demonstrated is your system not being entirely broken, but I don't see any success related to QA in your statistic. Ralf -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listin... -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, 2010-06-09 at 09:10 +0200, Ralf Corsepius wrote: On 06/09/2010 08:54 AM, Luke Macken wrote: On Wed, 2010-06-09 at 08:38 +0200, Kevin Kofler wrote: Luke Macken wrote: By success I mean that I felt we were successful in drafting, implementing, deploying, and utilizing the mentioned policies as expected, and the results show increased community engagement. This definition of success does not match mine nor the one you'll find in a dictionary. So your terminology is misleading. Really, Kevin? We're digressing to a dictionary battle? Fine, I'll play. First definition in the dictionary[0]: an event that accomplishes its intended purpose. Exactly. Your definition differs from Kevin's (and mine). Neither of you have mentioned your definition of the word success. Care to enlighten us? ...which is exactly what I meant to being with. To me, your definition of success is compliance with *your* process. If by your process you mean the processes created by the Fedora Community. I have had almost no say in the new updates criteria, nor am I on any rubber-stamping committee to approve it. I am, however, one of the *few* developers who actually works on bodhi, and I have a vested interested in improving it for the greater good of the community. Now, if the policies that are being approved do not actually benefit the greater good of the community, we have bigger problems. Whether this process is suitable to improve package quality, whether the technical system behind it is a good approach and whether your approach actually improves package quality or is mere bureaucray is highly questionable. Yes, all of those are highly questionable, with regard to this quality metric. To improve a process we must first observe how it is currently being utilized. What comes out of bodhi is what the maintainers put into it. Aside from that, we've been essentially been using it to croud-source QA. As expected, this is far from perfect, but it's a start until we have code in place that can perform rigorous and comprehensive testing. That said, all you demonstrated is your system not being entirely broken, but I don't see any success related to QA in your statistic. The numbers show an increase in community interaction. More eyes are looking at the updates and providing feedback. Considering we're croud-sourcing QA, an increase in participation is called a success. luke -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, Jun 9, 2010 at 8:31 AM, Adam Miller maxamill...@fedoraproject.org wrote: Did we really need to take some raw numbers that Luke was kind enough to put together and make it into some sort of QA methods holy war? The lesson here is that for data mining to make sense there must a consensus understanding of the question being asked. I'm not sure we have that here. I think the numbers are interesting in showing that bodhi can be used as a metric tool.. but I'm not sure we have consensus about what questions to be asking it. The strong reaction to the interpretive prose is just a symptom of that. Luke is asking bodhi questions about the trends of participation. I think some people have misread that as surrogate for a broader question about overall package quality. -jef -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On 06/09/2010 01:18 AM, Kevin Kofler wrote: If the packages have good quality, that means more testing is NOT needed, no matter what the actual amount of testing was. Apart from the Bodhi issue, I disagree with the logic of your statement. Quality doesn't exist (or at least is not provable) in an absence of testing; as the philosopher DeNiro says in Ronin, If there is any doubt, there is no doubt. On the contrary, there's the testing philosophy that says that if your tests start showing less faults, you should _increase_ and diversify testing, because you have simply exhausted previous test coverage profile. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, 09 Jun 2010 12:03:50 -0400, Luke wrote: According to the new acceptance critera, updates will have to spend some minimum amount of time in updates-testing, currently one week. Now, as to whether or not bodhi should auto-push after that week, that I'm not quite sure. Rest assured that you are NOT expected to auto-push anything unless packagers can opt-in or opt-out from such a process. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, 2010-06-08 at 16:51 -0400, Luke Macken wrote: I recently wrote some code to generate detailed statistics of Fedora EPEL updates within bodhi. Eventually this will be auto-generated and exposed within bodhi itself, but for now here are the initial metrics. This report definitely conveys the shortcomings in our testing, however, it does show us improving with each release. For Fedora 13, we implemented the No Frozen Rawhide process with improved Critical Path policies, which were definitely a success. With these enhanced procedures, along with the upcoming implementation of AutoQA and the new Package update acceptance criteria (https://fedoraproject.org/wiki/Package_update_acceptance_criteria), I think we'll see these numbers drastically improve in the future. You can find the code that generates these statistics here: https://fedorahosted.org/bodhi/browser/bodhi/tools/metrics.py https://fedorahosted.org/bodhi/browser/bodhi/tools/log_stats.py. If you have any ideas or suggestions for different types of metrics to generate, or if you find any bugs in my code, please let me know. Great stuff Luke, very interesting, and pleasing to see the improvements in the f13 process in cold hard numbers - thanks! -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org http://www.happyassassin.net -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Wed, 2010-06-09 at 09:51 +0200, Michael Schwendt wrote: from times when the fedora-easy-karma script wasn't available and didn't make it possible to mass-vote on updates. easy-karma doesn't allow you to mass vote, you still have to vote on each update individually. it simply streamlines the process somewhat, so you don't spend most of the time poking through the Bodhi web interface. just thought I'd note it... -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org http://www.happyassassin.net -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
Luke Macken wrote: Neither of you have mentioned your definition of the word success. Care to enlighten us? Success is the achievement of a worthwhile goal. If the original goal which was set is worthless, succeeding at it is meaningless. Now, if the policies that are being approved do not actually benefit the greater good of the community, we have bigger problems. Yet that's exactly the problem we're having. :-( We're being pushed towards more and more bureaucracy such as FORCED use of updates-testing without drawing any lessons from our project history: * Fedora Extras had no testing requirements, in fact it didn't even HAVE a testing repository. All builds were pushed directly to the stable repository. It worked great, to the point where Core merged with it. * Fedora Legacy had stringent QA requirements very similar to the ones which are about to be enforced now: updates could not move out of testing without either a minimum amount of positive feedback or a timeout (which had to be introduced because otherwise packages would never move out of testing). Feedback was counted separately for each distro version, just as our new policy will do it, which made it nearly impossible to get the required positive feedback for some releases. Over time, the amount of required positive feedback and the timeout had to be reduced several times because the system was just not working. In the end, Fedora Legacy failed, because it was impossible to deliver security updates in a timely manner with that kind of QA requirements. So why are we now going to use the Fedora Legacy model over the Fedora Extras one? Does FESCo really want Fedora to fail? Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
Adam Williamson wrote: Users also want regressions not to happen This is exactly why we need quick fixes, i.e. direct stable pushes: to be able to push a fixed update IMMEDIATELY if somebody caught a regression. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Thu, 2010-06-10 at 05:03 +0200, Kevin Kofler wrote: Adam Williamson wrote: Users also want regressions not to happen This is exactly why we need quick fixes, i.e. direct stable pushes: to be able to push a fixed update IMMEDIATELY if somebody caught a regression. We danced that tango last time, I'm not getting baited into it again. -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org http://www.happyassassin.net -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
bodhi statistics
I recently wrote some code to generate detailed statistics of Fedora EPEL updates within bodhi. Eventually this will be auto-generated and exposed within bodhi itself, but for now here are the initial metrics. This report definitely conveys the shortcomings in our testing, however, it does show us improving with each release. For Fedora 13, we implemented the No Frozen Rawhide process with improved Critical Path policies, which were definitely a success. With these enhanced procedures, along with the upcoming implementation of AutoQA and the new Package update acceptance criteria (https://fedoraproject.org/wiki/Package_update_acceptance_criteria), I think we'll see these numbers drastically improve in the future. You can find the code that generates these statistics here: https://fedorahosted.org/bodhi/browser/bodhi/tools/metrics.py https://fedorahosted.org/bodhi/browser/bodhi/tools/log_stats.py. If you have any ideas or suggestions for different types of metrics to generate, or if you find any bugs in my code, please let me know. luke = Bodhi Statistics Report (Generated on June 8th, 2010) = Out of 17412 total updates, 2958 received feedback (16.99%) Out of 1045 total unique karma submitters, the top 30 are: * notting (424) * mclasen (366) * jkeating (321) * adamwill (283) * cwickert (161) * rdieter (159) * pbrobinson (141) * kevin (141) * cweyl (122) * tomspur (119) * mtasaka (110) * xake (97) * cschwangler (86) * kwright (84) * peter (83) * hadess (80) * michich (72) * tagoh (69) * pfrields (69) * bpepple (69) * iarnell (68) * lkundrak (66) * shinobi (65) * sundaram (64) * spot (62) * pravins (62) * markmc (62) * thomasj (61) * smooge (60) * fab (59) Fedora 13 * 3562 updates * 3065 stable updates * 427 testing updates * 62 pending updates * 8 obsolete updates * 2371 bugfix updates (66.56%) * 745 enhancement updates (20.92%) * 89 security updates (2.50%) * 357 newpackage updates (10.02%) * 410 critical path updates (11.51%) * 1155 updates received feedback (32.43%) * 12120 +0 comments * 2477 +1 comments * 155 -1 comments * 595 unique authenticated karma submitters * 133 anonymous users gave feedback (1.57%) * 2261 out of 3562 updates went through testing (63.48%) * 1317 testing updates were pushed *without* karma (58.25%) * 21 critical path updates pushed *without* karma * Time spent in testing: * mean = 11 days * median = 9 days * mode = 7 days * 4 updates automatically unpushed due to karma (0.11%) * 0 of which were critical path updates * 231 updates automatically pushed due to karma (6.49%) * 2 of which were critical path updates * Time spent in testing of updates that were pushed by karma: * mean = 11 days * median = 7 days * mode = 7 days * Time spent in testing of updates that were unpushed by karma: * mean = 9 days * median = 5 days * mode = 5 days * 2445 packages updated (top 10 shown) * selinux-policy: 13 * jd: 12 * openoffice.org: 12 * gdb: 12 * ibus-pinyin: 11 * nautilus: 10 * kernel: 10 * evolution: 9 * libfm: 9 * libmx: 9 Fedora 12 * 4844 updates * 4291 stable updates * 371 testing updates * 113 pending updates * 69 obsolete updates * 2905 bugfix updates (59.97%) * 1054 enhancement updates (21.76%) * 201 security updates (4.15%) * 684 newpackage updates (14.12%) * 407 critical path updates (8.40%) * 960 updates received feedback (19.82%) * 16311 +0 comments * 1899 +1 comments * 554 -1 comments * 758 unique authenticated karma submitters * 576 anonymous users gave feedback (5.33%) * 2873 out of 4844 updates went through testing (59.31%) * 2138 testing updates were pushed *without* karma (74.42%) * 188 critical path updates pushed *without* karma * Time spent in testing: * mean = 14 days * median = 13 days * mode = 17 days * 12 updates automatically unpushed due to karma (0.25%) * 4 of which were critical path updates * 133 updates automatically pushed due to karma (2.75%) * 13 of which were critical path updates * Time spent in testing of updates that were pushed by karma: * mean = 11 days * median = 7 days * mode = 7 days * Time spent in testing of updates that were unpushed by karma: * mean = 9 days * median = 5 days * mode = 5 days * 2902 packages updated (top 10 shown) * qbittorrent: 25 * gdb: 25 * selinux-policy: 22 * kernel: 15 * xorg-x11-server: 14 * ibus: 13 * jd: 13 * abrt: 11 * gvfs: 11 * gtk2: 11
Re: bodhi statistics
Luke Macken wrote: This report definitely conveys the shortcomings in our testing, however, it does show us improving with each release. For Fedora 13, we implemented the No Frozen Rawhide process with improved Critical Path policies, which were definitely a success. With these enhanced procedures, along with the upcoming implementation of AutoQA and the new Package update acceptance criteria (https://fedoraproject.org/wiki/Package_update_acceptance_criteria), I think we'll see these numbers drastically improve in the future. Only because those numbers are taylored towards that very process (they measure the exact same things that process is going to enforce) and do not reflect the actual quality of the packages in any way. You can make really anything a success by measuring the very symptoms of the process and calling them a metric of quality. The reasons for which Bodhi karma (especially in its current incarnation) is a completely broken indicator of quality have been pointed out in several past threads. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, Jun 8, 2010 at 7:46 PM, Kevin Kofler kevin.kof...@chello.at wrote: Luke Macken wrote: This report definitely conveys the shortcomings in our testing, however, it does show us improving with each release. For Fedora 13, we implemented the No Frozen Rawhide process with improved Critical Path policies, which were definitely a success. With these enhanced procedures, along with the upcoming implementation of AutoQA and the new Package update acceptance criteria (https://fedoraproject.org/wiki/Package_update_acceptance_criteria), I think we'll see these numbers drastically improve in the future. Only because those numbers are taylored towards that very process (they measure the exact same things that process is going to enforce) and do not reflect the actual quality of the packages in any way. You can make really anything a success by measuring the very symptoms of the process and calling them a metric of quality. The reasons for which Bodhi karma (especially in its current incarnation) is a completely broken indicator of quality have been pointed out in several past threads. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel I'll have to agree with Kevin. I can't how any of those numbers represent the quality of anything. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, 2010-06-08 at 16:51 -0400, Luke Macken wrote: Fedora 13 * 3562 updates * 3065 stable updates * 427 testing updates * 62 pending updates * 8 obsolete updates Hey Luke, are these the numbers of F13 updates since we branched, or since the F13 release ? The numbers certainly look like the former, but I think the latter would be much more interesting. Matthias -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On Tue, Jun 8, 2010 at 6:41 PM, Brandon Lozza bran...@pwnage.ca wrote: On Tue, Jun 8, 2010 at 7:46 PM, Kevin Kofler kevin.kof...@chello.at wrote: Luke Macken wrote: This report definitely conveys the shortcomings in our testing, however, it does show us improving with each release. For Fedora 13, we implemented the No Frozen Rawhide process with improved Critical Path policies, which were definitely a success. With these enhanced procedures, along with the upcoming implementation of AutoQA and the new Package update acceptance criteria (https://fedoraproject.org/wiki/Package_update_acceptance_criteria), I think we'll see these numbers drastically improve in the future. Only because those numbers are taylored towards that very process (they measure the exact same things that process is going to enforce) and do not reflect the actual quality of the packages in any way. You can make really anything a success by measuring the very symptoms of the process and calling them a metric of quality. The reasons for which Bodhi karma (especially in its current incarnation) is a completely broken indicator of quality have been pointed out in several past threads. I'll have to agree with Kevin. I can't how any of those numbers represent the quality of anything. Well the only person I see mentioning quality is Kevin. And for some reason he is expecting it immediately when Luke said 3 things: 1) This was what he could initially see. 2) He thinks he sees improvement of what was measured, but he is interested in what would be better indicators. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel -- Stephen J Smoogen. “The core skill of innovators is error recovery, not failure avoidance.” Randy Nelson, President of Pixar University. We have a strategic plan. It's called doing things. — Herb Kelleher, founder Southwest Airlines -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
Re: bodhi statistics
On 06/09/2010 10:48 AM, Kevin Kofler wrote: Stephen John Smoogen wrote: Well the only person I see mentioning quality is Kevin. And for some reason he is expecting it immediately You can't claim that there are shortcomings in our testing (the exact words Luke used!) without a metric of quality. A shortcoming in testing means less testing was done than needed. If the packages have good quality, that means more testing is NOT needed, no matter what the actual amount of testing was. Packages without any testing is of unknown quality. If someone doesn't test them before pushing, users are unknowingly playing that role. Rahul -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel