GCC 3.0.4 is Out

Isle writes: "GCC 3.0.4 has finally been released. As those who has tried the prereleases will know this version finally compiles a working version of aRts and thus compiles the entire KDE-suite. With the Linux kernel compiling already with the 3.0.3 version, gcc 3.0 now compiles all major projects I know of. Is it finally time to dump that good old 2.95?"

Comments

C++ too slow

[Tersevs]

It is even slower at compiling C++ than the old one.
When will they implement precompiled headers.
Oh why, why. why havnt they done it already!
Or should i just give up Qt and move to GTK+?

Re:C++ too slow

[__past__]

How would using GTK+ make GCC use precompiled headers?

A more interesting question seems to be: Do the binaries run faster? Some people happen to run their apps more often than they compile them...

Re:C++ too slow

[neurojab]

There's one line in my environment setup in windows that makes it possible for me to compile my projects in the shortest possible time... DISABLE_PCH=T

Why do you want precompiled headers? Sure you might get a couple of faster compiles... when you're not scratching your head trying to figure out why PCH is failing.

GCC adding this feature would only cause me to spend more time compiling programs than I already do... unless they come up with a smarter implementation than MSVC.

Info propaganda

[__past__]

One "bug" I would really like to see fixed in GCC is the bogus "Linux and the GNU project" section in the info manual. Heck, it's supposed to be a compiler documentation, not a general "The world according to the FSF" rant.

What's next, the Emacs scratch buffer explaining why Free Software is better than Open Source? ls(1) warning about Non-GNU licensed binaries?

Re:Info propaganda

[__past__]

Actually, I am posting from a FreeBSD machine right now. That's where I read the GCC manual, what actually makes it even more annoying - not because I don't want anything GNU on my holy BSD (after all, FreeBSD does use their compiler - I'm fine with that), but because the Linux vs. GNU/Linux war isn't really relevant for me.

Also, I agree that Free Software is a better concept than Open Source, although the GNU approach isn't mine - for instance, they completely miss the freedom of the people writing the software and focus on the users only. However, there are better places to discuss such stuff than a compiler manual - the philosophy pages at GNU.org come to mind (and /. of course ;).

I respect the GNU approach, and wouldn't know any reason why they shouldn't point it out. But they developed an annoying practice of shouting pointless rants on quite unimportant details at inappropriate places. I'm sure most people reading the GCC manual know a bit about GNU, and if they care, about the GNU/Linux|Linux thingie. Can't we just get along and write code?

Re:Info propaganda

[__past__]

You completly misunderstood me.

I didn't intend to start a license flamewar. I didn't even mention the BSD license, and I didn't say anything about rights being denied writers - I just claim that the "freedom is a licensing issue" approach (regardless of which license, GPL od BSD or Apache etc) overlook the writers freedom.

I talk about a stylistic issue - even when your point is valid, you shouldn't annoy people. Insisting on silly details where totally inappropriate (and the insisting on GNU/Linux in the GCC manual is as appropriate as posting command-line parameters on the /. front-page) will do the Free Software community no good. That is my point, nothing else.

Oh, and if you would care to explain how exactly the BSD license leads to feudalism...

Re:C++ too slow, really!

[Tersevs]

GTK+ can be used in C that doesnt have support for templates.

I still believe that there is a limit of what compilation speed it acceptable. Have you compared different compilers? Yes, running speed is important too, but not at any price.

I find myself developing the software under Visual Studio and make sure that the program compiles and works there before moving to Linux and compiling the code. Id prefer to develop the stuff under Linux completly, but the time between typing make and having a binary just takes too darn long.
Sure i can have a cup of coffee during the compilations, but that sums upp to an awful lot of coffee...

By the way:
NUMBER_OF_USERS * BINARY_RUNNING_TIME_GAIN - NUMBER_OF_DEVELOPERS * NUMBER_OF_COMPILES * COMPILATION_SPEED_GAIN = A_NUMBER_THAT_I_DONT_CARE_ABOUT_IF_IT_IS_POSITIVE_ OR_NOT

glibc 2.2.5

[Skjellifetti]

Is also out now and according to the changelog, it finally works with gcc 3.

Re:Info

[Yarn]

I would be quite happy if the entire 'info' system dissapeared off the face of my installation, permanently.

It's a pita to use/write/index

man forever!

KDE speedup ?

[crsm]

A new 2.2.5 version of glibc is also out and this is perhaps even better news for KDE users. According to the glibc changelog:

• optimizations in the dynamic linker. binaries created by recent binutils versions start up quicker due to reduced time spend on relocations.

If I am reading this correctly, this is the much talked-about fix for the long startup times of KDE programs.

Yay :-)

Re:Info

[PD]

I use the --subnodes option and pipe through less to get a more usable printout. The info program has a really crappy interface.

like this:

info --subnodes bison | more

Does anyone know

[sinserve]

How much of ISO C++ is supported by the GCC compiler?
Not just the intel backen, but all the other popular ones (Sparc, PPC, etc.)

The only document I saw, is the BOOST library compile logs [note, boost.org is down at the moment, try a google cache.]

I know GCC beats the pants out of VisualC++ interms
of standard compliance, but how much better is it?

--

Re:Does anyone know

[Anonymous Coward]

There are a couple C++ features that nobody implements because they're too esoteric to put people on. There are a few C++ features that nobody implements because they're too hard to get into the compiler. Most compilers do the core set of features pretty well at this point. It's to the point where 95% of the spec and 99.99% of the spec that people actually use works just about anywhere.

Later versions of VS C have been getting more standards compliant for these core features. Check the documentation as VS C defaults to maintining backward compatibility (like K&R block variable scoping) over standards compliance in some respects. GCC has a few of the lesser used features that VS C does not. It shouldn't matter what backend you're compiling towards for supporting these features in GCC. VS C still does not do good templates although nobody does even great templates.

However, GCC thought it would be wonderful to add new features to C++ and so there are a bunch of programs locked into using GCC because they use GCC proprietary extensions. In that sense, I'd say GCC is much less standards compliant than VS C. If you are careful about avoiding the GCC extensions (or just disable them :), then it's probably a bit better than VS C but nothing out there is going to be much better.

By the way, you shouldn't try and use things like BOOST as a metric. As they say on their page:

"Warning: These tables are not a good indication of a particular compiler's compliance with the C++ Standard."

It's because a lot of code has either platform workarounds that hide bugs or proprietary extensions (like GCC's) that cause bugs on certain compilers.

Re:No "legal" closed-source C++ programs for Linux

[norwoodites]

The STL has an exception to the GPL:

// As a special exception, you may use this file as part of a free software
// library without restriction. Specifically, if other files instantiate
// templates or use macros or inline functions from this file, or you compile
// this file and link it with other files to produce an executable, this
// file does not by itself cause the resulting executable to be covered by
// the GNU General Public License. This exception does not however
// invalidate any other reasons why the executable file might be covered by
// the GNU General Public License.

So you can use any license for C++ programs.

We tried removing it once...

[devphil]

...but RMS won't let us.

There are days when I dream about another GCC fork.

Been there, done that

[devphil]

It is even slower at compiling C++ than the old one.

That's because C++ got bigger since the old one.

When will they implement precompiled headers.

A precompiled headers branch was created some time ago. Feel free to try using it. It's been implemented by two different commercial groups already; they're just merging in one of those solutions.

Perhaps you think implementing precompiled headers is easy? I invite you to try.

Oh why, why. why havnt they done it already!

If are are dissatisfied with the compiler, and unwilling to contribute your own time to make it better, you should definitely demand a refund of the money you paid for it.