The fact that it stayed there for a week is even stranger. Nobody tested the code for a whole week?
Well, Micheal did test it. The regression has been found and is being fixed now. All of that long before any release. So the system works, right?
He should have actually reported it right away but as I understand it chose to make a little experiment to see if anybody else will find it. Nobody did in a weeks time, but there is still lots of time until the code is ready for a release, so it doesn't necessarily mean a disaster.
Everybody should test before commiting but a single contributor cant test for all configurations. One have to rely on others to get all of the possible combinations tested. And look: it worked right away!
Performance regressions are especially hard to find since most test frameworks don't measure it.
Obviously kernel developers should use more performance testing frameworks such as phoromatic but maybe they are relying on the automatic kernel tracker at phoromatic.com to alert them of just those kind of problems.
Probably not a lot of people. Many people waits to -rc1 (or even later) to start testing things. It turns out that there is not a huge "disaster" anywhere, during the development of a release there're always petitions to reverse commits that cause problems.
6) Testing is hard, regression management is hard, it places constraints. A lot of the kernel developers simply don't care for some of these constraints. I can guarantee that if this was raised on the kernel mailing list, there would be a lot of people not focusing on the issue, but mouthing of on the testing, the methodology, etc. ie: Attacking everything but the issue.
Look how poorly they reacted when the regression was raised.
That was an independent statement regarding an issue that had been found. At this stage, it is unclear if this is the same regression (via a private thread). Or even the train of analysis is correct for the issue Michael raised.
Identifying a regression (even just pointing to the Phoromatic tracker) would have had the usual mix of "the benchmark is pointless", "it's ubuntu, not the kernel" sorts of discussions. I've been through it a few times already.
well, one problem with ubuntu: since there are no kernel hackers employed by Canonical, there are no kernel hackers using ubuntu. If you have a problem while using Fedora, opensuse etc the chances are a lot better.
And when reporting problems with gentoo I always got either no reaction at all - or patches to try out
1. This article contains useful information that it's nice to see publicized.
2. It would have been nice to see more in-depth analysis of the cause, but this is by no means necessary since the information is useful on it's own. Top tier websites like Ars or Anantech would have gone deeper into the details, but I understand that those websites also have a lot more resources than Phoronix.
3. Most of the people are only complaining because of the title. Change that, and you have no problems. The issue is that the title reads like a tabloid - what disaster did Michael even find? If this regression was released in the final 2.6.35 kernel, that would be a disaster. A regression during the merge window that can easily just be backed out before even RC1 hits? Not a disaster, not even unexpected. The title seems intended solely to drive page views by promoting sensationalism, which is why most people are critical. Judging by the number of comments here, it seems like it may have succeeded, though, so I suppose we'll be seeing more of these in the future.