Hacker News Reader

Comment by pie_flavor | Hacker News Reader

Stallman is an ideologue, in every sense of the word; a crusader for free software. His worldview can sometimes be so small you can't fit your head inside it. That sometimes means he makes management decisions about Emacs optimizing for being a word processor, when nobody else in open source cares about word processors. That also sometimes means he makes management decisions about GTK optimizing for more accessibility, when nobody else in open source cares about accessibility. You can't have the good without the bad. If RMS were to be replaced, there would be no more high profile sillinesses, but at the same time, a huge number of things would slowly get worse because he's the only one who cares enough about them.

View on HN

By high profile silliness do you mean arguably piloting free software into the ground?

GCC is almost literally nonexistent as a place for new compiler research in large part thanks to previous stallmanist approaches to managing, this isn't just some cutesy hypothetical about principles and so on.

The above also ignores stallman's manner around people.

> GCC is almost literally nonexistent

I believe that you are referring to the old LLVM modularity? That is old news. GCC is now (reportedly¹) better in C++20 standard compliance, and this will probably not change, since, (reportedly²) both Google and Apple have moved away from LLVM involvement.

1. https://news.ycombinator.com/item?id=30653685

2. https://news.ycombinator.com/item?id=32924033

Not particularly. I don't really care about C++ compliance but rather fundamental new algorithms and optimizations in the backend.

When I read papers on new compiler techniques it's never GCC getting the implementation, this saddens me but it's true.

The points I made in the reply to the first link still stand, Although things are actually improving.

> [...] this will probably not change, since, (reportedly) both Google and Apple have moved away from LLVM involvement.

Correction: according to what you've linked, it is not LLVM but clang. This correction would be extremely important for my reply.

I can't see how is that a good news for GCC. This (and other incidents) suggests that both companies have lost interest in investing the future of C++, and they were already building their own languages for a long time. You can argue that they wanted to bend the standard and implementations as they see fit---and it's partly true, otherwise why would you want to join at all---but it's probably more likely that C++ is becoming irrelevant at least in some applications. And if the language is made irrelevant, a better standard compliance means nothing.

Perhaps not, but even if true, this was not, then, any fault of GCC, which was the topic at hand.

Note that the topic in hand i.e. mentioned in my assertion is compiler research

> Google and Apple have moved away from LLVM involvement

LLVM is very much alive at all big tech, to my best knowledge. It's C++ that's dying there (and I say - good riddance).

Exactly, all the later languages that were made to replace C++ (Swift, Rust, Zig) all rely on LLVM, so it has become a much more important project than in the past. (And I'll bet Google's new Carbon language will also rely on LLVM.) It's just that it's so hard and time-consuming to write a compiler backend with all the latest optimization techniques, and GCC's backend isn't as much as flexible than what LLVM brings to the table.

For compiler research LLVM has a good external API and it is easy to add in new passes without having to fight the rest of the code-base.

Reply on news.ycombinator.com