GCC 4.1 Released

Luineancaion writes "Looks like GCC 4.1 has been released. From what I know this includes the GNU Classpath merge and means that Azureus can now be used in a 100% Free-Software system. Thanks to everyone that worked on it, and keep up the good work!"

Comments

Know and love GCC

[JoeShmoe950]

As a developer, I love GCC. Its great, easy, and best of all free. GCC is probably one of the most benifical open source projects around, more important even than linux.

Re:Know and love GCC

[Leroy_Brown242]

I see them more as a symbiosis.

Where would GCC be without Linux?
Where would Linux be without GCC?

Re:Know and love GCC

[chthon]

GCC existed long before Linux, but Linux made it possible to deploy it large scale on PC's (I know about DJGCC).

When I got into PC's in school in the 80's, Borland ruled.

When I started working in 1990, I wanted to learn C++. There were no inexpensive options however. I knew about GNU and FSF, but all the software they sold was targeted at commercial 32-bit platforms (I think that they did not even had a C++ compiler back then). All other options where too expensive, until the summer of 1991, when Zortech came out with their C++ compiler, which included also 286 and 386 extension software, which made it possible to run 32-bit code.

Linux a year later made it possible for people really interested to run real 32-bit code on a real 32-bit platform.

But you should really add another question to your list :

Where would both be without the Internet ?

Re:Know and love GCC

[DrXym]

In other words, GCC would be exactly where it is today, had it not been for Linux.

I doubt that. GCC was seriously stagnated way before 2.95/3.0 (hence the reason egcs appeared for a while) and was no match at all compared to various commercial compilers. Linux was about the only popular OS which *needed* a modern gcc and thus most of the development came from Linux stakeholders - Red Hat etc. Without Linux I fully expect that the compiler would be an also-ran by now, along with most commercial Unices.

on the java side

[petermgreen]

are they using the gcj as chacheing jit (e.g. GCJ run on demand to turn class files into shared objects which are then loaded dynamically) system that was mentioned in one of the papers i read recently or what?

Java status?

[harmonica]

From the story: From what I know this includes the GNU Classpath merge and means that Azureus can now be used in a 100% Free-Software system.

Sounds interesting. Is there any ChangeLog to read? I browsed the gcc and the gcj pages, but I couldn't find anything.

Changelog?

[Theovon]

No mention of a changelog? If you're going to announce something, it sure would be nice to have a link to a page that explains some interesting stuff about what's new in it. I've tried looking at their wiki, but its 'news' section and its stuff on 4.1 hasn't been updated since like March.

Sense and portability

[noz]

I was always angry with Sun touting Java(R)(TM)*** as portable when run-time environments were made available for only a small (albeit popular) set of architecture/operating system pairs. My Alpha running Debian at home and my Alpha running FreeBSD at work were left cold, lonely, and wanting Java; running a subset of Java applications with free software partial implementations. This is a triumph for FOSS.

I'm thinking of contributing to GCC...

[TwoBit]

I love GCC, but I lament that its ability to do inlining is rather bad.
I'm wondering how hard it would be join the project and work on rectifying this.

Not a programmer but

[Anonymous Coward]

what (if any) implications will this have on OO.o?

LLVM and GCC will likely merge within a year

[Anonymous Coward]

It looks like GCC will merge with LLVM. Follow the discussion thread to see key GCC developers agree to the many benefits of using LLVM. And even if LLVM is not folded into GCC, Apple plans to use LLVM as its backend anyway.

Re:Home depot

[Billly Gates]

I saw some benchmarks a few months ago that closes the gap in performance in c/c++ performance with gcc/g++ 4.0 and the Intel compilers.

Intel wrote them when gcc2.95 was still out and c++ performance was not that good nor was it truly modern ansi compliant by the iso. For example things like the STL were merely emulated and performance for non x86 cpu's was behind too.

With gcc3.x and now gcc4.x its fully caught up in almost all area's. Its nice now to have a nice c/c++ compiler for the alpha and mips processors that produce fast code.

I wonder if the rise of Linux and Free software is what made the compilers catch up?

By the way the Intel compiler is still the way to go for Fortran.

Re:Actually, it's not released yet

[Matt Perry]

I'm not sure what gave the person who submitted the story, or the editor who posted the story, the idea that 4.1 was released, but it isn't.

The GCC home page very clearly states "Current release series: GCC 4.1.0" while below it all the others say "Previous release series" and one at the bottom says "Active development (mainline)." That indicates as clearly as can be which are the old, the current, and the development releases.

If you are right, and 4.1 wasn't released, then the GCC web page might need to be reworded to better communicate that fact.

Re:C and Objective-C

[mark-t]

Hmmm... the textbook I used in the compiler course I took way back in the olden days said that recursive descent compilers were, in general, slower and in general larger than state-based ones.

Mind you, the book is over 20 years old now.

GCC website

[Per Abrahamsen]

It is surprisingly hard to find out what the current release is from the GCC webpages. The front page has a misleading 4.1. Press "Releases", and you get a misleading 3.4.4. You have to go on to the "Development plan", under Future timeline, to find the actual latest and greatest 4.0.2 somewhere down the ASCII art tree.

Re:C and Objective-C

[am 2k]

According to a lecture I took last year, this has changed in recent years. Now it's much more important to work on a small set of data at a time, since the processors have large caches. The recursive approach helps to keep the data you're working on inside the cache.

20 years ago nobody had a cache, so it was an imperative to keep your stack size down.

GCC is important, but what about progress in C++?

[master_p]

GCC is a very important piece of open source software, but it is time to see some real progress in C++:

1) type inference and the 'auto' keyword, as it was suggested a while back.

2) real garbage collection. It is not possible to have effective collection without support from the compiler. The Boehm's collector is not a precise collector, because it can not use type information; it is only the compiler that knows such things.

3) a standard library that goes beyond collections, algorithms and files (and is based around garbage collection, mentioned above). We need gui, threads, sockets, XML, etc. Java did it, I see no reason why C++ can not do it. The argument that it is too difficult to do it no longer holds, due to the cross platform toolkits that exist (but none of them is standard or open as GCC).

And I don't want to hear anyone say 'why don't you join GCC to do it yourself', because I have a full-time job working as a C++ programmer and I have barely the time to take a bath, plus it is extremely difficult to enter the GCC development process, due to being highly not-documented (you have to read the sources) and pretty much a closed circle.

Re:GCC is the Key to Open Source's Success

[Anonymous Coward]

Wrong. There's an install process called stage 1/3 that produces the same result as a stage 1 installation using a stage 3 binary installation. The only other noteworthy source-based distro, SourceMage, does exactly the same: you install binaries and then rebuild the system.
Oh I almost forgot: don't go around spreading FUD you stupid piece of rotting bat shit

Re:Home depot

[macshit]

As for the compiler itself, yes, it's not the best of the bunch, but that doesn't mean it's any bad either (quite the opposite!). Intels' compiler, for example, still beats it for performance (at least the last few times i tried it), but i could live happily with GCC alone.

I've spent quite a bit of time hacking on gcc, and I'd say my biggest complaint is that a lot of the gcc code really sucks. It's chock full of gigantic impossible-to-understand chunks of code -- few comments, huge numbers of global variables, an "enumerate every case I could think of with 25 page if-statements" coding style, vast numbers of unwritten assumptions about the way your processor works. That it works at all, never mind as well as it does, is a testament to the dedication of gcc hackers.

I think a lot of this is historic, and the newer parts of the compiler are much better (and so the overall code quality is slowly improving as old code gets replaced), but gcc can still be a real pain to work on. If you're trying to port to an architecture that differs in some way from "typical" architectures, be prepared for misery.

Re:GCC is NOT open source

[wootest]

All political discussion aside, isn't being Open Source a prerequisite of being Free Software (and to avoid stepping on toes: no, that does not mean that I think Open Source came before Free Software)? I think GCC is both Open Source and Free Software, simply because it needs to be Open Source if it's going to be Free Software; releasing your code and allowing others to fork it are key parts of both Open Source and Free Software. I guess the point I'm trying to make is that it's not *just* Open Source (which would imply not being Free Software), but it is by definition *also* Open Source.

Re:Masters of understatement

[IamTheRealMike]

No, IIRC this is called Mudflap and generates calls to special glibc APIs which check buffer sizes for calls like read(). It's another layer of security added into Linux lately: on top of exec-shield and SELinux things are feeling pretty secure round here ... too bad not every distro has these things (basically on Red Hat distros have them all).

Re:That's good.

[slavemowgli]

Is it? How do you know? Maybe it's just me, but I've uploaded about 150 GB in the past 7 months on BT on my measly 192 kbps uplink, and not a *single* bit of that was infringing on copyright. Why do people always think that BT doesn't have any uses besides copyright infringement? It's not true, and while it *seems* that copyright infringement accounts for a significant part of BT use, there simply is no data that would show that it's actually the #1 use (of course, there's no data that shows it's not the case, either, but without any data at all whatsoever, you shouldn't pull numbers out of your ass).

Re:C and Objective-C

[Anonymous Coward]

Part of the problem with state based parsers is that reduction rules can cascade quite a bit for each input token, especially with a complex grammar. It's possible to optimize away a lot of the reductions by just noting that whatever rule you're evaluating can only reduce to another rule, etc. and simply apply the reduction once. I imagine this is the biggest reason that recursive decent is faster in practice. It probably also makes generating the syntax tree easier, since the structure of the parse tree can be used to implement the recursion, making everything faster. No need to keep track of both a state based parser and a syntax tree. I haven't looked at gcc's parser though, so I don't know exactly what it does. It's probably smarter than me.