Torvalds Has It With "-Wstringop-overflow" On GCC Due To Kernel Breakage

[phoronix]

Phoronix: Torvalds Has It With "-Wstringop-overflow" On GCC Due To Kernel Breakage

One of the new features for Linux 6.8 that was merged late was enabling the -Wstringop-overflow compiler option to warn about possible buffer overflows in cases where the compiler can detect such possible overflows at compile-time. While it's nice in theory, issues on GCC has led Linus Torvalds to disabling this compiler option as of now Linux 6.8...

https://www.phoronix.com/news/Linux-Drop-GCC--Wstringop-of

[aschmidtm]

I thought that said wingstop

[timofonic]

I think Linux kernel code needs to be a lot more hardened and very actively checked with zillions of tools automatically in a behemoth infrastructure. Shit like this wouldn't happen so often.

Will progressive Rustification help on this? Just a bit, but the infrastructure needs to be massively improved too. Even Git needs to learn new tricks.

[timofonic]

I thought that said wingstop

Those overflows have very different effects!

[F.Ultra]

I think Linux kernel code needs to be a lot more hardened and very actively checked with zillions of tools automatically in a behemoth infrastructure. Shit like this wouldn't happen so often.

Will progressive Rustification help on this? Just a bit, but the infrastructure needs to be massively improved too. Even Git needs to learn new tricks.

this isn't them finding new real buffer overflows in the kernel, this is one of the "actively checked with zillions of tools automatically" tools being broken in their reporting.

[dEnigma]

I think Linux kernel code needs to be a lot more hardened and very actively checked with zillions of tools automatically in a behemoth infrastructure. Shit like this wouldn't happen so often.

This is about the tools doing the checking possibly being broken (and apparently definitely being broken in the case of GCC version 11)

[JEBjames]

@Michael

Typo

"disabling this compiler option as of now Linux 6.8." maybe "now for Linux"

[bregma]

I think Linux kernel code needs to be a lot more hardened and very actively checked with zillions of tools automatically in a behemoth infrastructure. Shit like this wouldn't happen so often.

Will progressive Rustification help on this? Just a bit, but the infrastructure needs to be massively improved too. Even Git needs to learn new tricks.

This is one of the new false positives coming from the GCC static type checker. If you can prove the Rust toolchain will never have bugs in any of its analyzers (like the borrow checker, for example) then sure, switching to something new will probably git rid of the old bugs.

[Weasel]

I think Linux kernel code needs to be a lot more hardened and very actively checked with zillions of tools automatically in a behemoth infrastructure. Shit like this wouldn't happen so often.

Will progressive Rustification help on this? Just a bit, but the infrastructure needs to be massively improved too. Even Git needs to learn new tricks.

You realize this entire issue exists because the "zillions of tools" report it wrong? Yes, those tools have bugs. Who would have though, right?

They're warnings for a reason, not errors. I will never understand people who turn warnings into errors by default. It's OK when you want to experiment but turning it on for production is beyond idiotic, especially since it varies by compilers and bugs.

[Volta]

I think Linux kernel code needs to be a lot more hardened and very actively checked with zillions of tools automatically in a behemoth infrastructure. Shit like this wouldn't happen so often.

Will progressive Rustification help on this? Just a bit, but the infrastructure needs to be massively improved too. Even Git needs to learn new tricks.

You sometimes sound like a bot.