Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Linus Torvalds on regressions

lkml.org

Post reply on HN
Print view View original

11–19 of 19 posts

Re: Linus Torvalds on regressions

#11

Post by Forge36 »

pimeys wrote: post #5
Yuioup wrote: post #3

Earlier quoted context omitted.

_

Did we read the same text? I didn't read it as exploding, but just as a teacher teaching the pupils about the rules of kernel and what not to brake. The text was very clear and informational.

I read it as exploding. The post above yours was deleted when I saw yours, so I won't defend it, but share hire I read Linus's update.

I believe the exploding appearance was an intentional writing style. We reading it because a) it's Linus and b) he very clearly communicates his position to the point of being crass. It works and is stated in the opening: "Dammit. I haven't had to shout and curse at people for a while, but this is ABSOLUTELY THE MOST IMPORTANT THING IN THE UNIVERSE WHEN IT COMES TO SOFTWARE DEVELOPMENT"

Makes a clear statement AND expresses a clear tone. Was he teaching? Yes. Did he make his point? Yes. Was he calm? Behind the keyboard maybe. But in the email I'd wager he didn't want to look calm

Re: Linus Torvalds on regressions

#13

Post by dagw »

zaarn wrote: post #2

There is much less swearing going on than in the last few rants. Usually that means Linus is extra serious. edit: More interestingly, he later explains his behaviour (and sort of apologizes for exploding): https://lkml.org/lkml/2018/8/3/727

There is much less swearing going on than in the last few rants. Usually that means Linus is extra serious.

I remember reading a chef (might have been Bourdain) who said something like when the head chef is shouting "Where the fuck is the fucking lamb you useless piece of shit" then things where basically on track, but if he calmly said "can you please hurry up with the starters for table 6 and 7" then he knew he was in deep trouble and really had to pick up his pace.

Re: Linus Torvalds on regressions

#14

Post by buth_lika »

axilmar wrote: post #12

I like him!!!! even if I don't agree with him sometimes. People need to be a lot more straightforward in their daily dealings with others. The world would be a much better place.

I agree.. people can be dishonest and hurtful while being perfectly polite, just like they can be honest, kind and caring while swearing like sailors. That is the difference that matters. When you ignore that and just look at the form, you get https://en.wikipedia.org/wiki/Perverse_incentive

Here's Linux talking about being straightforward, and why he cannot afford ambiguity:

https://www.youtube.com/watch?v=MShbP3OpASA&t=35m51s

Re: Linus Torvalds on regressions

#17

Post by craftyguy »

> We do not regress

This made me laugh, then cry. I regularly hit kernel regressions in functionality (e.g. worked and now doesn't) in both my personal and professional/work systems. Sadly, the components are not niche drivers, it's things like ath9k, i915, iwlwifi, alsa (and several drivers there), etc. The kernel desperately needs more testing than just a handful of developers dog fooding it before calling it 'good.'

Re: Linus Torvalds on regressions

#18

Post by elygre »

craftyguy wrote: post #17

> We do not regress This made me laugh, then cry. I regularly hit kernel regressions in functionality (e.g. worked and now doesn't) in both my personal and professional/work systems. Sadly, the components are not niche drivers, it's things like ath9k, i915, iwlwifi, alsa (and several drivers there), etc. The kernel desperately needs more testing than just a handful of developers dog fooding it before calling it 'good…

Also, if you find a kernel regression, be sure not to create a workaround in your userland application. The gods might get angry with you.

Re: Linus Torvalds on regressions

#19

Post by TooBrokeToBeg »

I don't understand why this is so interesting, as it doesn't seem to be anything about regression per se.

A userland tool supported functionality that it can't anymore, because it isn't available in a new kernel. This is not surprising, as people find ways to access and change data outside of published APIs all the time and they get patched away. The userland author said that the kernel upgrade would break old versions of the tool. So what?

Post reply on HN