Chrome now uses clang for production builds on Linux

http://article.gmane.org/gmane.comp.compilers.clang.devel/39700

by espadrine • 12 years ago

169 159 12 years ago

article.gmane.org

tedks8• 12 years ago

It's been a good run, but we're finally seeing the end of free compilers.

Over time, using Clang effectively will require proprietary modules from Google, Apple, et all and the GCC project, long dead, will haunt us from the grave.

We'll tell stories to our children about how when we started learning to program, it was easy to get started because compilers were free, but now the only free compilers are toys, ill-suited for anything but the most basic education.

A special thanks to everyone who helped make this possible by attacking GCC and championing LLVM whenever possible. I hope it's very, very easy for you to write extra modules for your new proprietary compiler.

kev009• 12 years ago

Yeah running FreeBSD on all my systems really depletes my wallet too.

mahouse• 12 years ago

I'm sorry, but I don't understand. Aren't both Clang and LLVM open source?

chimeracoder2• 12 years ago

OP is being a rather hyperbolic, but it's not really that unreasonable to criticize Clang/LLVM on this issue.

Both projects are politically motivated. GCC began explicitly as a project to provide a FOSS C compiler (few if any existed at the time). Clang/LLVM didn't start at Apple, but their development has been heavily sponsored by Apple, whose interest in the project stems from their interest in excising the GPL (and specifically the GPLv3) from their entire development stack. (Notice how the version of bash that ships with OS X is almost a decade old, and is also the last version that was released under GPLv2).

The fact that Clang/LLVM are released under a FOSS license today does not mean (A) that they always will be, or (B) all essential functionality will always be[0]. With GCC, being a FOSS compiler collection is their entire raison d'etre - it's right in the name![1]

Edit: Reworded sentence to remove ambiguous antecedent.

[0] (ie, it's conceivable in the future that nobody will want to use Clang/GCC on its own, but will essentially have to use it in conjunction with another piece of related software, and there's no guarantee that this software will be FOSS. With GCC this is not a relevant concern only because the FSF would never let that happen; it undermines the entire founding point of the project.

[1] GNU Compiler Collection

tedks8• 12 years ago

Not completely. Proprietary modules and forks of both exist, so in the broadest sense they are not (in the sense of, if I'm ever handed a copy of something calling itself LLVM, I do not know that it is libre software).

But worse than that, proprietary modules can be mixed-in to Clang and LLVM. So essential functionality can be hidden in proprietary modules. Imagine that Apple wants to do something nifty for Swift, but doesn't want Google to be able to use it in Go -- they'll release a proprietary module that can be linked into Clang.

All free code in LLVM can be forked and made proprietary, and no proprietary code can be reasonably expected to be liberated. So the balance of free to non-free code can only ever get worse.

I'm sure that the libre software world being based on a proprietary compile will be a great thing to add into the mix of pervasive surveillance and network penetration.

sneak• 12 years ago

Stop this nonsense. Not everything that isn't exactly according to Stallman's vision is the devil. Black and white thinking helps no one.

http://llvm.org/docs/DeveloperPolicy.html#license

To others: I presume what the parent comment is complaining about is that companies are legally allowed to modify the llvm source and distribute those binaries without source code.

This is not a bad thing, unless you're an open source zealot.

From the URL above:

We intend to keep LLVM perpetually open source and to use a liberal open source license.

and

We believe this fosters the widest adoption of LLVM because it allows commercial products to be derived from LLVM with few restrictions and without a requirement for making any derived works also open source (i.e. LLVM’s license is not a “copyleft” license like the GPL).

This is reasonable.

cliffbean• 12 years ago

The current leaders of the LLVM project intend to keep it open.

They have also traded away a way to encourage others to keep it open in return for wider adoption.

Interpret that as you will.

nashashmi• 12 years ago

Your reply and others' reply makes me wish I could read the post. This [flagkilled] is just complete nonsense.

Igglyboo3• 12 years ago

LLVM has a far more "free" license than gcc does. GPLv3 is indeed less "free" than MIT style licenses regardless of what RMS says.

chimeracoder2• 12 years ago

LLVM has a far more "free" license than gcc does.

The Clang/LLVM license is both a free license as defined by the FSF and an open source license as defined by the OSI[1].

[1] https://en.wikipedia.org/wiki/University_of_Illinois/NCSA_Op...

voidlogic2• 12 years ago

That depends on if you are encouraging "free" for societies interests or "free" for individual/corp interests. GPLv3 encourages the former, but not the later. The two interests almost always come into some degree of conflict (although in many ways are compatible).

I view the flamewarz between BSDish and GPLvX licenses as similar in spirit to the arguments between laissez faire capitalism and northern European social capitalism.

yarrel• 12 years ago

Assertions without evidence etc. ;-)

M2Ys4U• 12 years ago

It's not about being more or less free. You should ask the question "freedom for whom?"

People who use Copyleft licenses want the users of the software to have freedom. People who use permissive licenses want the developers of the software to have freedom.

tedks8• 12 years ago

By this logic a dictatorship is the freest form of government. No restrictions!

TallGuyShort• 12 years ago

Remember that time that Apache destroyed software freedom because it wasn't GPL? Me neither.

angersock• 12 years ago

Yeah, and nobody was ever able to move to another web server again, and now we're stuck with Apache forever. Forever.

idontknow123• 12 years ago

If you have a competitive advantage why contribute back in any meaningful way? We have seen this type of stuff on Android with Dalvik (where Qualcomm has a special library / modified version of the VM).

That's not to say it destroys freedom, but its not ideal for a certain subset of software. The GPL is also not magical in this regard, since you have projects where things don't get up streamed.

bhouston• 12 years ago

I think Clang could see the end of the MSVC compiler except where .NET compatibility is required. Thus this is actually a win for open source software. Seriously, Clang is just awesome in terms of performance, debugging ability and standard support compared to MSVC.

Corrado• 12 years ago

I wonder, with the recent open sourcing of C#/.Net, would a Clang/LLVM module for them be out of the question? That would make for some sweet ability to port applications around. :)

qwerta• 12 years ago

Funny, after some exposure to GNU codebase (including GCC) I used to thought, that evil M$ is trying to destroy free software, by poluting it with horrible code.

blinkingled4• 12 years ago

I am not entirely sure why you got downvoted so much - probably due to the 'attacking GCC' part which I disagree with you on, but for what it's worth I agree with your point that proprietary modules and forks are a risk with precedence.

For instance Apple came out with A8X which I am sure required special clang codegen optimizations - where is that code? If that continues to sit behind Apple's closed walls then it's already proving OP's point.

ajross3• 12 years ago

Exactly. The lack of Apple's release of its LLVM/clang code for its popular platform is a problem. And it might be a reasonable argument to make about copyleft in the abstract (c.f. Apple's very different behavior toward WebKit which was enforced by license).

But it's not an argument against clang, which remains a high quality free software compiler and one worth using. And citing the existence of said free software as evidence for the "end" of free compilers seems... weird.

DannyBee19• 12 years ago

I own the C++ language platforms team at Google (which includes the LLVM and GCC developers).

Google literally upstreams everything we produce for llvm/clang.

We did the same for GCC, including things we could have kept to ourselves, like LIPO (https://gcc.gnu.org/wiki/LightweightIpo)

I have a lot of trouble coming up with a case where we wouldn't upstream something. Do you have any evidence for your concern, based on history, or is it all random hyperbole?

As both a GCC and LLVM developer, i can tell you that GCC lost entirely of its own accord.

PS it's somewhat ironic that Google contributed a lot to GCC, and you complain about it and assert they have poor motives because they decided to move to a different technical solution, yet from what i can tell looking at changelogs, you have never contributed to either GCC or LLVM. If you want the world to turn out the way you want it, you actually have to participate.

diminish2• 12 years ago

GP's concern of proprietary modules seem valid IMHO. Based on your track record and plans ahead can you confirm this is not a risk?

tedks8• 12 years ago

I don't think Google has poor motives at all. This is the way the wind is blowing. Google is almost certainly right to move to LLVM and you'd be far better situated to comment than I am.

Personally, I see something as big as Chrome moving to LLVM as a late-stage indicator of the mass adoption of the platform. I'm young, as you've pointed out, and I've never worked for Google, but I'd bet that this decision wasn't made lightly, and it certainly wasn't made early in the lifetime of clang/LLVM.

The people truly at fault for the things I fear, should they come to pass at all, are not the late adopters like Google, but the innovators who decided to put their energy into LLVM, knowing full well how dangerous it could be.

I don't and didn't mean to imply that Google is going to stop contributing, or that Google never has.

I do think that Google's decisions to contribute upstream are based on business sense, as opposed to morality, and I think that should the day come that it makes business sense for Google to not contribute back, or to release a non-free fork, it'll do that. As others in this subthread have pointed out, Apple has already distributed non-free LLVM modules.

The concern I have is that this is possible with LLVM, and it was not possible with GCC. When Apple wrote a non-free Objective-C frontend for GCC, they liberated it because the Free Software Foundation successfully enforced the GPL. There is no legal mechanism to force corporations to contribute back to LLVM, so they do it of their own goodwill, if they do it at all.

I think this is dangerous because it can lead to a world where there are no good free compilers. The people who wrote LLVM, the people who championed it early on and contributed to it, they could have realized this too, and they did nothing to stop it. I truly hope they acted the way they did because I am egregiously wrong, and that nothing I'm afraid of will ever happen.

To put things in perspective, GCC had one fork in its history already. I think that if it was forked again, redesigned extensively to make it easier to get parse trees and add modules and so on, with the legal requirement that all such modules be free in any distributed compiler, the FSF would eventually, hopefully, accept that as the canonical GCC. It would have been harder to do that than to make LLVM, maybe it would have cost the world a few years of compilers research, but it would have been possible, and it would have made for a more secure outcome.

kasabali2• 12 years ago

You get it all wrong. He is not saying "Google will not contribute back to LLVM" per se. He said "Any entity can leverage LLVM without contributing back. It's very nice Google is playing fair with the community, but not all commercial entities will be that friendly, and LLVM's licence makes it easier. He is not attacking Google and there is no reason to take it "personally" and entering to the "defensive mode" immediately.

rakoo• 12 years ago

Does anyone have details on why they made the switch ? The email states very little improvements, even a bug on 32-bit Debians. Is this only because of the license of the compiler ?

CJefferson3• 12 years ago

They do mention:

With the switch to clang, binary size dropped about 8% while performance mostly stayed the same. Startup time improved a bit,

Also I can imagine it might be nice to use the same compiler on different OSes, and gcc is very much a 2nd class citizen on mac.

moyix• 12 years ago

On the other hand, clang is very much a second class citizen on Windows.

IshKebab• 12 years ago

The licence of the compiler doesn't affect what you build with it. I expect the reasons are the same most people are switching to clang:

* Modern modular design (everything is libraries) * Faster compile times * Nicer error messages

The modular design in particular is a good reason that might not be obvious at first. It lets you do things like debuggers, code formatters and static analysers in a much more sane and robust way than GCC's "no plugins, everything must be separate and flaky" way.

pjmlp• 12 years ago

Clang brought new life into C and C++ eco-systems, specially in the sense that since we cannot get rid of C for the time being, at least we can improve the tooling around it to make it safer.

But sometimes I wonder if the license doesn't play a role as well.

deng6• 12 years ago

GCC has support for plugins since version 4.5, which is over four years old. Also, the error messages have improved greatly in the last releases.

film42• 12 years ago

I recently heard a google employee come speak at my university about how google's main codebase is compiled using clang, and their janitorial work is done using clang mapreduce [1]. Furthermore, that google employee talked about how google is working with the LLVM team to integrate the toolchain into visual studio for cross-platform compilation [2]. Pretty cool stuff!

[1] (similar talk) https://www.youtube.com/watch?v=mVbDzTM21BQ

[2] http://llvm.org/docs/GettingStartedVS.html

joelthelion• 12 years ago

None of these reasons matter at all for production builds. The production build needs to happen only once per release, and the only things that count are performance and reliability.

The last time I checked gcc still had better performance than CLang, but things might have changed since then.

emn132• 12 years ago

I have no idea whether it's true, but on the topic of gcc's inferiority in the matter of error messages, this source claims it's no longer true: https://gcc.gnu.org/wiki/ClangDiagnosticsComparison

Of course, given the source, apply salt :-).

nly• 12 years ago

Google are invested in Clang by virtue of employing guys like Chandler[0] to work on it. When you're scrubbing up monstrous codebases like Chrome, tooling matters. Clangs architecture, and things like all the sanitizers that are part of the broader project, presumably go a long way.

[0] https://twitter.com/chandlerc1024

mcescalante2• 12 years ago

I posted a link to a 2011 talk from Chromium devs about using clang in another comment here, but below is a summary of what they seemed to enjoy from clang (at least, from the talk points):

- Useful warnings, they look good, and are on by default.

- More than a compiler, lets you build your own tools (example: AddressSanitizer)

- Better performance (less time) with clang vs. gcc for both -O0 -g and -O2 (at the time the benchmark was done on 8 cores, 24GB)

In summary / tl;dr : they enjoy diagnostics, the clang code base, how it works, and the ability to write their own tools. It seems like it's more to do with preference/performance and less to do with licensing. I'm sure that now they're doing many more things with clang since three years has passed since the aforementioned talk.

DannyBee19• 12 years ago

Because you can get a single toolchain that is state of the art, can build for all of chrome's supported platforms (in fact, it could do it from a single binary if you really wanted), and one can easily improve to get the performance you want out of.

amelius2• 12 years ago

How about Chromium? Are they switching too?

mcescalante2• 12 years ago

AFIAK, Chromium core devs are all Google employees, and their code makes up the base for Chrome, although Chrome is Google branded and includes other features such as usage tracking. I found a talk given in 2011 by two Chromium devs about using clang, which seems to point to nearly the same increase in build performance along with a few other reasons they like using clang. Anyways, seems like they've been using clang for a while. http://llvm.org/devmtg/2011-11/Weber_Wennborg_UsingClangInCh...

metzman• 12 years ago

Chromium doesn't ship binaries to end users. Linux distros that have chromium build it from source. But I suspect that they are probably going to switch to clang as the recommend compiler for development.

amelius2• 12 years ago

Well, in any case they have a build system in place, I hope :)

Which either defaults to using clang or gcc.

pix642• 12 years ago

They are here.

https://commondatastorage.googleapis.com/chromium-browser-sn...

i_am_ralpht• 12 years ago

Next stop Windows! Then they can use all the modern C++ constructs they like...

Chrome now uses clang for production builds on Linux

Related Stories