Hacker News Reader

Comment by h2odragon | Hacker News Reader

Torvalds was a strong advocate of GCC 2.95 (iirc), early on in Linux history, because he knew the kind of code it would emit and didn't trust the newer compilers to produce code that was correct in those circumstances.

The workarounds and effort required to tell a compiler today that no, you really did want to do the thing you said might well be insupportable. I figure they started going astray about the time self modifying code became frowned upon.

View on HN

To be fair, the backend in the early GCC 3.x series was just kind of stupid sometimes. Even now I find strange if cheap and harmless heisengremlins in GCC-produced x86 code (like MOV R, S; MOV S, R; MOV R, S) from time to time, while the Clang output, even if not always good, is at least reasonable. This is not to diss the GCC team—the effort required to port a whole compiler to a completely new IR with a completely different organizing principle while keeping it working most of that time boggles the mind, frankly. But the result does occasionally behave in weird ways.

Can Linux compile under clang nowadays?

Yes, see: https://www.kernel.org/doc/html/latest/kbuild/llvm.html

While sibling comment is correct that the simple answer is "yes", I'd like to add that Google uses Clang as the preferred compiler for Android. There's a little bit of drift between Android's version of Linux and mainline, but the effect is still that building Linux with Cland is being extensively used and tested.

Yes, when that Linux variant is called Android.

Since almost 5 years by now.

Reply on news.ycombinator.com