Kees: The condition we need to catch is the case of unexpected wrap-around.
Linus: If you go "wrap-around is wrong" as some kind of general; rule, I'm going to ignore you, and I'm going to tell people to ignore you, and refuse any idiotic patches that are the result of such idiotic rules.
Kees: I would phrase the rule as "ambiguous intent of wrap-around is wrong".
Linus: Any kind of mindless "this can wrap around" is inexcusable.
@dentangle@kees I had a similar argument once about ring buffers and maintaining the end of it as one past the last element, because it makes things easier (TL;DR), and I only managed to shut the other person up by sending a photo of the appropriate page in the Art of Computer Programming.
I hate appeal to authority, but it can work wonders sometimes.
Gotta say.. just Wow! What a behavior this shows. Literally stamping on an idea. Bringing full authority to bear and minimize any chance of people more broadly starting to muse on merits and possible drawbacks to take into account. Absolutely toxic stuff.
@jens@smallcircles@dentangle You'd think so! But honestly this is a giant improvement. Only one threat and only one veiled ad hominem attack. There was no swearing nor direct insults. 🙄 So, yeah, it's better but still pretty off-putting for most people. I remain sad about all the developers we'll never see join the community.
Having made it painfully clear you are not talking about UB or objecting to intentional overflow, to then be repeatedly clubbed with that one thing no matter how many times you correct the record...
It's wilful. There's no way your meaning was missed.
@dentangle Convincing someone about some technical things is hard enough. Having them threaten, "I will use my position of power and influence to specifically ruin your attempts to be heard", though, is the troublesome bit. If I'm so wrong, present the evidence. Don't threaten to silence me. Sheesh.
@dentangle@kees strange, I didn't see anything extreme in his responses.
Kees makes the valid point that in order to really catch any bugs, you must take an allowlist approach. By definition then, you are making all wrap-around "bad" by default, and only allowing special case exceptions.
Linus says no, we can't forbid by default something that C programmers all know to be normal behavior. He then highlights more specific cases that could be targeted. What's unreasonable about that?
@hyc@kees If Linus had made his point as you just did, sticking to the technical points, nothing. There's a valid debate to have.
But he didn't.
Linus SHOUTED IN ALL CAPS, described such proposals as "idiotic", implying that Kees is therefore an idiot, and kept referring to a point that Kees has never made.
All Kees is proposing is clarifying programmer intent to prevent a known, frequent and recurring class of errors found in the kernel. He has never said that wraparound is "bad".
Add comment