GCC 4 [LWN.net]

GCC 4

Posted Jan 17, 2005 20:21 UTC (Mon) by steven97 (guest, #2702)
In reply to: GCC 4 by danielos
Parent article: Fedora Core 4 plans announced

What makes you think this? I can assure you, I'm quite happy with GCC 4 as it is developing now. Sure, there are some problems here and there, but nothing big really. SUSE, RedHat, and Apple are all already using GCC 4 derived compilers for serious work, and the number of serious bugs is going down each day.

I agree that GCC 4.1 will have fewer rough edges than GCC 4.0, but that is what you get with a x.0 release. But if you think about this as a distribution builder, the last thing you want is to release a GCC 3 based compiler in your beta distribution now. If that beta is the base distro that you build your enterprise release on, then you risk ending up supporting an old compiler infrastructure for the next five years. GCC 4.0 is radically different internally from GCC 3, so you do not want to have to do that. And Fedora is of course, in a way, a RHEL-beta-like product. So going with GCC 4.0 makes a lot of sense for Fedora.

(Log in to post comments)

GCC 4

Posted Jan 17, 2005 21:46 UTC (Mon) by yokem_55 (subscriber, #10498) [Link]

In addition, it is easier for binary based distro's to use the latest and greatest comilers since workarounds that are needed for building various packages are easily masked from everyone using those binaries. For source driven distro's like Gentoo however, the quality of the compiler is essential to the everyday health of the system, as everything is build by the user and unless the workarounds are already integrated into the build scripts, a newer compiler can wreak absolute havoc. Hence, gentoo, which is for the most part a bleeding edge distro, still uses the gcc-3.3 line of compilers by default for the 32-bit x86 arch.

GCC 4

Posted Jan 22, 2005 12:42 UTC (Sat) by danielos (subscriber, #6053) [Link]

Something user (not programmer) use to do is compiling the kernel, and there are a number of regression in this field that make gcc 3.3 better (and 2.95 better than 3.3). But this do not matter: as you said it's a beta.

What really could make think break is C++ changes that could break source compatibility: a lot of good work have be done it this area but there are some bug outstanding that, IMO, should be fixed for 4.0.0 release and are postponed for 4.1 instead. I repeat, IMO this policy is not very good, also, given the good work gcc developer are doing in these day for bug fixing, I guess that fix some parser bugs is not a so hard work. But steering comite state this changes are not possible now.

I use 4.0 myself, I don't see a lot advance. GCC 4.0 is the base architecture for a better optimization but a small part of this is done.