GCC 12.2 Compiler Released With 70+ Bug Fixes

[phoronix]

Phoronix: GCC 12.2 Compiler Released With 70+ Bug Fixes

Out this morning is version 12.2 of the GNU Compiler Collection...

https://www.phoronix.com/news/GCC-12.2-Compiler-Released

[rene]

incl. major x86 stack usage bug reported by Linus Torvalds himself. Updated: https://www.youtube.com/watch?v=nvbJ3kCFmQY

[ClosedSource]

Unless you need newer C++ features or specific compiler features, it's safer to wait till gcc XX.2 or XX.3 releases anyway.

[timofonic]

What about Rust in GCC? Any news?

[Michael]

What about Rust in GCC? Any news?

That's all planned for GCC 13 (or later), not GCC 12.x that is just stable bug fixes now.

[Malsabku]

How is it possible that something as important as a compiler has so many bugs?

[timofonic]

How is it possible that something as important as a compiler has so many bugs?

Why not? Compilers are very difficult to develop, specially with so many broad platform and programming language support. A lot more than LLVM world in both aspects.

They detect the bugs, that's a lot better than "no bugs at all".

[arQon]

How is it possible that something as important as a compiler has so many bugs?

Because modern compilers are immensely complex. There are secondary factors too, but that's the main one by a long way.

70 isn't actually all that bad.

[NateHubbard]

How is it possible that something as important as a compiler has so many bugs?

As opposed to what software that doesn't have bugs?

[Xake]

incl. major x86 stack usage bug reported by Linus Torvalds himself. Updated: https://www.youtube.com/watch?v=nvbJ3kCFmQY

I actually read https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105930 and liked it quite much. Much more so when you look at the behaviour of the persons involved.
Linus does much job to investigate as much as he can and is not directly insulting, even when he feels he does it in an ugly way, admits it because he has problems reading the docs, and also somewhat blaming the formats of the docs for why he might not find a better way to bisect, suggesting ways to document it he might finds more useful but without trying to demand it.
He also finds a possible culprit in a commit that when reverting removes the problem he is seeing, but also leaves it open that just reverting this commit may not be the way to fix the issue, and that even before he later comes to the conclusion that the commit he found might not really be the culprit, but something that exposes a deeper issue after further investigations and testings of what other suggests.
And that is also a thing. He tests what is suggested as workarounds from the compiler developers, discussing what happens when following the suggestions, helping trying to figure out more efficient ways to reproduce to lock down on what goes wrong so the proper fix can be found.

Yes he knows a lot more than many users does about how what his software does, what a compiler is supposed to do, and from it might do more on his own that regular users. But the interactions here is still something some regular users should learn from. Many bugreports I see gets nearly the same treatment from other developers/maintainers in trying to figure out what goes wrong even if the user has not nearly the same amount of experience as Linus. That is as long as the users are genuinely trying to do their best to help the developer, and not start to go haywire in for example the kernel-bugzilla and because their favorite binary blob stopped working demanding reverts of stuff they know nothing about.

Many maintainers of open source software are helpful even if you are not Linus Torvalds, as long as you are as helpful in trying to figure out what goes wrong and testing suggestions as Linus is in this bug-report.