LLVM Replaces libstdc++ Library With libc++

[RealNC]

clang is great for developers. GCC is great for users. Both are currently very valuable.

[XorEaxEax]

So it may be a hard thing to transition away from GCC, but clang has a higher upside. GCC will continue to get messier and harder to maintain while clang hopefully has a good 25 years before it accumulates that much cruft.

Not true, since GCC's code base is continously being enhanced and thus getting less 'messy' and less 'harder to maintain'. Again I don't understand those who wants one compiler to emerge 'victorious', that's the worst possible outcome. Lack of competition ALWAYS leads to stagnation. The best of worlds is GCC and Clang continously going head to head, driving their developers to create the best possible products. The only reason I can see for people to want one compiler to fail is if they have some external agendas. Also it would seem alot of these people are not even using the compilers in question judging by their poor knowledge regarding them.

[Nobu]

What's more fun though is trying to speculate on whether or not software would be further along than it is today had there never been any controlling laws to begin with. I believe it would be, since even if there were more attempts at keeping code closed, leaks of that code would bust everything open and the amount of source code circulating throughout the public would be much more advanced as everything could much more easily be built upon. Money would be generated purely by services and bounties (paid development sprints), what I believe will eventually be the future and most likely have to take place before it is finally realized that licensing laws deprive everyone of a better quality of life due to the needless wasteful money spent in courtrooms that it causes. At least until money itself is abolished.

So, basically, it'd be like your mom's secret recipe for meatloaf, which she once told you and which you "accidentally" slipped to one of your friends, etc..

[theoddone33]

You expect Clang to support (and be stable) on 20 architectures in 2 years?

I don't, no. I expect it to support and be stable on the handful of commonly used architectures though.

However, in a couple of years it could be the case that when companies come up with new architectures (like the Cell, for instance) they base their compilers around LLVM instead of GCC, and I would love to see that happen.

I believe (perhaps incorrectly) that most of the work on GCC is done by engineers at companies like that, though. So it may be a hard thing to transition away from GCC, but clang has a higher upside. GCC will continue to get messier and harder to maintain while clang hopefully has a good 25 years before it accumulates that much cruft.

[pingufunkybeat]

Clang is not there yet and LLVM is still playing catch up a little, but at this stage both projects are very promising. In 2 years I hope we're at the point where GCC can be ditched altogether and serve only as a relic of a time when it was OK to develop software that's extremely hard to maintain.

You expect Clang to support (and be stable) on 20 architectures in 2 years?

Or maybe you misunderstand the whole point of GCC?

Though I do agree that code maintainability is an issue, and one that is not easily addressed.

I'm aware, but you missed my point.

I'll give him credit for starting Linux off, but my point was that his views on communism are flawed because Linux can be copied on demand, but physical items cannot.

Your point is that RMS is a crackpot, and you refuse to use his stuff because you are waging a holy war on communism?

Well, good for you, I guess. He showed that the Free Software model works very well for software, and it's not something that Novell or IBM or Intel or Microsoft would have done. Although they are paying developers today.

RMS forced them to pay developers to create Free Software (through the GPL). And that's quite an achievement.

[smitty3268]

It should make things easier to manage, because everything is strictly modularized. Performance has historically been an issue, but there are no recent benchmarks. Performance is not always a deciding factor. A microkernel is particularly useful for environments where reliability matters. It's difficult to bring down a microkernel system, because a crash / infinite loop is localized to its module, which is then restarted. Plus, it's really small. If you have limited resources, it's a good fit, too.

Good for a real time OS as well, since it's much easier to make guarantees.

However, realistically, it would be a huge amount of work to replace everything the Linux kernel does with another one. And the result would be slower. So without a good reason (like scalability in the hundreds of cores cpus that may be coming soon) I don't see it happening for a mainstream desktop OS.

[Remco]

@NoEffex



I heard microkernel makes things even harder to do, more bloated and it works slower... Know a single, good microkernel?

It should make things easier to manage, because everything is strictly modularized. Performance has historically been an issue, but there are no recent benchmarks. Performance is not always a deciding factor. A microkernel is particularly useful for environments where reliability matters. It's difficult to bring down a microkernel system, because a crash / infinite loop is localized to its module, which is then restarted. Plus, it's really small. If you have limited resources, it's a good fit, too.

[bridgman]

How about QNX ?

[kraftman]

@NoEffex

TBH with modern hardware I think a microkernel or a hybrid kernel would help keep things not so bloated, but that's just me (There's a LOT of room for improvising with micro/hybrid kernels).

I heard microkernel makes things even harder to do, more bloated and it works slower... Know a single, good microkernel?

[kraftman]

GCC is awefully bloated. Any changes that want to happen tend to be rejected. Same with the Linux kernel. All good things must come to an end..

This is a total and idiotic bull when comes to Linux. You want to dictate what shouldn't be rejected? You can cut off whatever part you want at any time and devs can do the same. You can consider it's bloated as a whole package, so every architecture, file system etc. etc. makes a big package with a lot of code, but till there are enough devs, maintainers everything is fine and under control. There will be problem if Linux would be indivisible, but it's not. GCC is bloated (as I read even at lkml), but it's much more mature, it's faster, it supports far more architectures then llvm, so it's not going to die very soon (if ever, if it will also stay under control).