Community
Participate
Working Groups
The following link [1] shows there are a few problems which have 0 as the number of reporters. This looks confusing because there is at least one reporter who caused the first incident to be logged. I think it would be better to show a positive number there. [1] https://dev.eclipse.org/recommenders/committers/confess/#/problems/?page=0&size=10&sort=numberOfReporters,asc
True. I've recognized them too. They are a result of the weekend's deduplication run and have no visible influence on the current system. I'll investigate those corner cases later.
(In reply to Marcel Bruch from comment #1) > True. I've recognized them too. They are a result of the weekend's > deduplication run and have no visible influence on the current system. I'll > investigate those corner cases later. Thanks! Another deviation that I observed is that today the number of incidents is about 404k, but on Friday it was about 450k. Similarly for problems, it is about 10k but was over 13k. I am not sure what it means, but the database seems to be a bit smaller after deduplication ;-) I don't have any hard data to prove it so I can't file a reliable bug, but I was quite surprised to see the numbers are lower today.
(In reply to Szymon Ptaszkiewicz from comment #2) > (In reply to Marcel Bruch from comment #1) > > True. I've recognized them too. They are a result of the weekend's > > deduplication run and have no visible influence on the current system. I'll > > investigate those corner cases later. > > Thanks! > > Another deviation that I observed is that today the number of incidents is > about 404k, but on Friday it was about 450k. God damn! You noticed almost every weekend change :) You are right. I did a data cleanup and removed exact identical error reports (same user, same trace, sample bundles, same comments). I also removed error reports larger than 300 KB and introduced a filtering that does not accept larger elements anymore. The size cleanup removed a few hundred error reports. So the majority comes from exact similar reports filtering. I wonder whether I should create a log of all changes. Maybe I should post such updates on the EPP forum so that everyone can subscribe to these...
(In reply to Marcel Bruch from comment #3) > God damn! You noticed almost every weekend change :) Almost every? Challenge accepted! ;-) > I wonder whether I should create a log of all changes. Maybe I should post > such updates on the EPP forum so that everyone can subscribe to these... +1 for mailing list, and since it potentially affects all projects I wouldn't mind using cross-project list for that.
Created attachment 255757 [details] Screenshot (In reply to Szymon Ptaszkiewicz from comment #4) > +1 for mailing list, and since it potentially affects all projects I > wouldn't mind using cross-project list for that. I think cross-projects is not the right place since: i) no all committers are subscribed to that list ii) eclipse users certainly won't find any information about aeri there. iii) I'd feel like a spammer to send all kinds of updates and information to that list. I've discussed that earlier with Markus and we came to the conclusion that the EPP forum would be an appropriate alternative to cross-projects. I just started posting notes in that forum. Please note that you can easily subscribe to that forum - just click on the "Subscribe" label on top of the forum page. You need to be logged in though.
https://www.eclipse.org/forums/index.php/f/69/
(In reply to Marcel Bruch from comment #1) > True. I've recognized them too. They are a result of the weekend's > deduplication run and have no visible influence on the current system. I'll > investigate those corner cases later. Just to document what I found, it seems that the number of problems which have 0 reporters is growing and even some newly reported ones ("Created On" says "15-09-25") have 0 reporters. This means that the deduplication did not change only the existing problems but also affects how the number of reporters is calculated for new problems.
Just to give an update: We are down to 62 out of 30,500 reports with zero reporters. It looks that the duplicate detection for OutOfMemoryErrors and StackOverflowErrors still has some problems.
Good news, there is only 1 problem with 0 reporters right now.
Right. For that one the initial incident was removed during an over-eager cleanup operation. The problem is kept since it's tracked in Bugzilla. I can live with that one off-by-one problem, and thus, close this bug now.
The bug is back again: https://dev.eclipse.org/recommenders/committers/aeri/v2/#!/problems/?c:!%28%28t:empty%29%29,sort:!%28%28property:numberOfReporters,direction:ASCENDING%29%29
Not a bug this time - but a misleading ui. Incidents are removed when they are older than 6 months and so are problems that did not receive any traffic in the past 6 month. Those that are assigned to a bug, however stay in the system but with zero "recent reporters". I'll need to change the presentation in this regard.
The Eclipse AERI server and components have been decommissioned in 2019, see Bug 551898. Therefore any remaining bugs are now being closed as WONTFIX. If this bug is still relevant, please make a comment and we'll move it to the correct project/component for further investigation. This change was made as part of a bulk change.
