Linux 3.0 Release Delayed

JustinRLynn writes "A recent Google+ Post by Linus Torvalds indicates that version 3.0 of the Linux kernel will have to wait due to the discovery of a 'subtle pathname lookup bug.' Linus indicates, 'We have a patch, we understand the problem, and it looks ObviouslyCorrect(tm), but I don't think I want to release 3.0 just a couple of hours after applying it.'"

Comments

Chicken?

[oldhack]

Release it now, you fool!

Re:Chicken?

[X0563511]

Project Manager: Did it build?
Developer: Yea, but we haven't even run the thing yet
Project Manager: Ship it!

Re:Chicken?

[NoNonAlphaCharsHere]

Your project managers make you get a completely clean build before you ship? How do you guys stay on schedule?

Re:Chicken?

[arth1]

Your project managers make you get a completely clean build before you ship? How do you guys stay on schedule?

Simple enough, Firefox style: Any time you get a semi-clean build, you tag it. When you release, you simply bump the version number of the last tagged build. So what if you don't get half the features - it's a new version, as witnessed by the version number!

Re:Chicken?

[Canberra Bob]

I for one am glad a geek IS the project manager.

More focused on getting it right than meeting deadlines.

That would explain why GNU/Hurd is such an excellent OS

Re:Chicken?

[shentino]

RMS isn't a geek. He's a fanatic.

Re:Chicken?

[leenks]

In a decent engineering environment, regardless of it being "agile" or not, I would hope that the decision was made to formally release to customers, and a branch was created for that release. After a short period of stabilisation (removal of critical bugs, but no more features) the most recent build from that branch can be released, while feature development has continued on master/trunk. All of those bug fixes can be merged back into master for mainline development.

Sure, you might provide canary/integration/every-stable-build-of-master-from-the-CI-system releases to stakeholders so they can track development and give feedback, but your releases need to be managed. How else are you ever going to support that release after your customers have got hold of it - ie supply bugfixes for it without forcing them to upgrade to a newer release that might contain masses of new features, bugs, and incompatible changes?

Path names? Bah.

[MrEricSir]

I say push it live, let those damn n00bs grow some chest hair by referencing all their files by inode id.

Re:Path names? Bah.

[Anonymous Coward]

Fuck you: a gang of inode ids killed my father.

Re:Path names? Bah.

[Anonymous Coward]

This is GNU/Linux we are talking about right? So people actually wait for bugs to be fixed before pushing a release forward? That's news to me.

Re:Path names? Bah.

[dudpixel]

here's hoping said n00bs aren't female...

Re:Path names? Bah.

[NotBorg]

No. The topic is actually just Linux. Also, it's not uncommon for Linus to hold off on a release if things aren't "quiet" enough.

Re:Path names? Bah.

[NoNonAlphaCharsHere]

You too? I had a special character in my nick, and my old account has been broken for several months. Hence, the new name.

Re:Path names? Bah.

[aynoknman]

My name is Inode Montoya. You killed my father. Prepare to die.

Re:Path names? Bah.

[NoNonAlphaCharsHere]

I hate the silent ones. Especially in elevators.

Re:Path names? Bah.

[rubycodez]

inode id: no Luke, *I* am your father!

Re:Path names? Bah.

[DarkOx]

My name is Inode Montoya, You unlinked my father. Prepare to be free()'d.

Perhaps today IS a good day to die

[RogueWarrior65]

LET'S SHIP IT!!!

No problem.

[loxosceles]

No problem. I'll just run GNU Hurd.

HOW DARE THEY

[DWMorse]

The shareholders will demand an answer for th ... wait.

Google+ is still in testing too

[cervesaebraciator]

Waiting for someone to take this as Linus Torvalds' recommendation of Google+...

Re:Google+ is still in testing too

[jampola]

Okay, that's the last time I will reply to anyone semi-famous in G+ where they are sure to have about 100 replies thereafter! Especially since G+ decides it wants to subscribe my email to the thread and email me each reply! Seriously, what the fuck is with that??!!

Re:Google+ is still in testing too

[Tubal-Cain]

He opted in by commenting on the post. Keeping everyone that is a part of a discussion "in the loop" is consistent with the purpose of social networking.

It isn't Google's fault that jampola chose to reply to a post which received a high signal-to-noise ratio.

Re:If it's a patch applied to 3.0

[adamstew]

No, only firefox does that.

Google Plus

[Wizarth]

I think the more important news here is - Linus uses Google+ for announcements now?

Facebook really is in trouble now.

Re:Google Plus

[shutdown -p now]

Google should really start advertising G+ as "Facebook, but with people you actually care about".

Re:Google Plus

[Zebedeu]

Makes sense to me. Unlike Facebook, G+ allows anyone with an account to follow Linus's public posts without him having to accept them as his "friends".

It's perfect for this type of announcement. It's Twitter for those who felt constrained by the character limit.

Re:Google Plus

[Zebedeu]

You can look at a profile without needing an account. For an example, try Linus' own: https://plus.google.com/102150693225130002912/posts

However, there doesn't seem to be an RSS feed (though I could be mistaken), so I don't know how you'd follow his posts without visiting his profile often.

Fair Enough

[jampola]

Reasons like " I don't think I want to release 3.0 just a couple of hours after applying it" would never fly in a commercial environment. This is why I love Linux and pretty much anything open source. I know it's cliche but whaddyagunnado?

Re:Fair Enough

[Anonymous Coward]

I don't know what commercial environment you've been in, but in the places I've worked, release becomes hell because you have your bug list and someone (read a commitee) has gone through and labeled the "show stoppers" which are bugs deemed important enough to be fixed before the software can be released, and because of politics in the commitee, all but the most trivial become show stoppers. Upon fixing the last show stopper, the software then needs to go through regression at a minimum, and usually a complete test suite before it's allowed to be released. And even then, that goes into system integration, where the whole process starts again.

Re:Fair Enough

[ATMAvatar]

That depends on the balance of power in the organization. If sales/marketing have the bigger share of the power, QA is downsized or eliminated and the only "show stoppers" are unchecked feature boxes.

Re:But wait, what is this OS i'm using now?

[besalope]

You have a release candidate. The official launch of RTM has been delayed.

Re:If it's a patch applied to 3.0

[GNUALMAFUERTE]

You are confused. This is the Kernel, not Firefox.

BTW: I heard the guys at Mozilla are working on a new feature: The ability to change the version number while the browser is running. That's real progress.

Hollywood celebrities and such forth...

[wyoung76]

... all moving to G+ and posting there instead will be the death of Facebook, but not a moment before

Re:If it's a patch applied to 3.0

[renegadesx]

Yes but that feature has been delayed until the release of Firefox 7, so you will have to wait a week.

Here's what the bug was!

[Theovon]

Sadly, I don't understand the explanation or what the patch changes.

https://lkml.org/lkml/2011/7/17/103

Re:Here's what the bug was!

[FlyingGuy]

Ok, from my reading of the patch which could be WAY the fuck wrong BTW, I think it is a race condition between the unlinking of a file and returning the inode to the pool AND the CP command ( copy a file ) traversing the inode list. In other words the CP command was trying to stat a file that was partially unlinked do to the update of the node list still being in progress.

If you still don't understand that don't feel bad, I had to read and re-read the note like 10 times before I probably got this explanation wrong.

Re:Here's what the bug was!

[jez9999]

The solution to this is really Obviously Correct; I don't know why they didn't post the bug in the summary:

That -ENOENT in walk_component: isn't it assuming we found a negative
dentry, before reaching the read_seqcount_retry which complete_walk
(or nameidata_drop_rcu_last before 3.0) would use to confirm a successful
lookup? And can't memory pressure prune a dentry, coming to dentry_kill
which __d_drops to unhash before dentry_iput resets d_inode to NULL, but
the dentry_rcuwalk_barrier between those is ineffective if the other end
ignores the seqcount?

Here's what the bug was!

[FrootLoops]

https://lkml.org/lkml/2011/7/17/103

[Posted by Theovon earlier, but I prefer a clickable link.]

Re:The emperor isn't wearing any clothes

[macshit]

It can take time for some people to get used to git ("wait the commands aren't exactly the same as CVS!? noooooooooo...").

But once it clicks, you'll never want to go back.

There's a good reason git is by far the most popular "new generation" source-control system (and no, it's not "because Linus is popular"). It's simply more powerful, more facile, more nimble than the competition.

Re:If it's a patch applied to 3.0

[shutdown -p now]

BTW: I heard the guys at Mozilla are working on a new feature: The ability to change the version number while the browser is running. That's real progress.

That is true. They still do require you to restart the browser after that happens, however, so that extensions would properly stop working as they should.

Re:The emperor isn't wearing any clothes

[Serious Callers Only]

For everything besides committing, Git is horrible

It would be nice to know what you had a problem with. People here could perhaps enlighten you as to why things aren't working out for you, or you could enlighten them as to why git is inferior. It has its flaws (chiefly obscure error messages), but I've found it a better fit than cvs and svn.

Re:"+1 Sad but true"

[TaoPhoenix]

There is! It's 'Underrated"!

Re:Linus Torvalds and Google+

[maxume]

Recently.