GCC 4.5 Is Still Not Ready For Release

[phoronix]

Phoronix: GCC 4.5 Is Still Not Ready For Release

GCC 4.4.0 was released nearly a year ago, but it looks like its one-year anniversary may pass without a new major release of the GNU Compiler Collection. GCC 4.5 was not yet branched back in January due to outstanding P1 regressions, which is also blocking any release candidates from being made available. Now in March there are 16 regressions of P1 status still outstanding...

http://www.phoronix.com/vr.php?view=ODA2Ng

[SheeEttin]

Richard Guenther pleads with the GCC developers to work on any assigned P1 regressions or un-assign yourself from them, otherwise those regressions will be downgraded to P2 so that this new release can be made.

Well, that's one way to eliminate high-priority bugs... :\

[deanjo]

Well, that's one way to eliminate high-priority bugs... :\

Sadly this seems to be a growing trend now days and isn't isolated to GCC.

[FireBurn]

Though to be fair at least they do have a "it's released when it's ready" mantra rather than rushing utter tosh out the door to meet a schedule

[RealNC]

I hope we won't end up at some point closing everything as WORKSFORME just to make a release :P

[V!NCENT]

Though to be fair at least they do have a "it's released when it's ready" mantra rather than rushing utter tosh out the door to meet a schedule

Better a release than no release at all, or when your too old to still operate a computer...

If these regressions are not functionality but speed, then just release it (other speed improvements will make up for it).

If these regressions are feature cripple than don't release it.

But if the devs aren't doing anything at all, then releasing it may make them get their asses in gear. Humans need motivation, you know.

PS: Debian is for the people that want stable and bug-free software. You can't possibly have an excuse

[Shining Arcanine]

With GCC 4.5 the MPC library has been integrated to evaluate complex arithmetic at compile time more accurately

The MPC library is a fork of the GMP library, which was integrated with GCC 4.4. It is like someone copy and pasted the GCC 4.4 documentation and replaced GMP with MPC in it, but regardless, this change was more about helping MinGW ports of GCC to Windows than anything else, because one of the reasons the GMP library was forked is primarily because its developers were hostile to having it run on Windows. Another was its switch from the LGPL to the GPL. Hypothetically speaking, bug fixes that were included in GMP under the GPL cannot be included in MPC unless their authors release them under the LGPL license, so I am not sure if this is an improvement for GCC or not.

Though to be fair at least they do have a "it's released when it's ready" mantra rather than rushing utter tosh out the door to meet a schedule

When did "those regressions will be downgraded to P2 so that this new release can be made" become "it's released when it's ready"? This seems Microsoft's approach to Windows Vista development, and we all know how that turned out.

I would rather see GCC 4.5 take another year to become stable (possibly with a GCC 4.6 developed with better optimizations in parallel) than to see it to be released with P1 bugs that were downgraded because they were unable to fix them in time for a deadline.

[FireBurn]

The MPC library is a fork of the GMP library, which was integrated with GCC 4.4. It is like someone copy and pasted the GCC 4.4 documentation and replaced GMP with MPC in it, but regardless, this change was more about helping MinGW ports of GCC to Windows than anything else, because one of the reasons the GMP library was forked is primarily because its developers were hostile to having it run on Windows. Another was its switch from the LGPL to the GPL. Hypothetically speaking, bug fixes that were included in GMP under the GPL cannot be included in MPC unless their authors release them under the LGPL license, so I am not sure if this is an improvement for GCC or not.



When did "those regressions will be downgraded to P2 so that this new release can be made" become "it's released when it's ready"? This seems Microsoft's approach to Windows Vista development, and we all know how that turned out.

I would rather see GCC 4.5 take another year to become stable (possibly with a GCC 4.6 developed with better optimizations in parallel) than to see it to be released with P1 bugs that were downgraded because they were unable to fix them in time for a deadline.

They still have release rules so just down grading the bugs to P2 doesn't mean they can release, there would also have to be fewer than 100 bugs in total.

This isn't any different from any previous GCC release where the same has happened

The P1 bugs are mostly ICE (internal compiler errors) and GCC not building on what should be a primary target.

[V!NCENT]

Compile errors o.O

GLHF... If the 4.5 release ships with such severe bugs it won't be included in any distro but Fedora...

[FireBurn]

Compile errors o.O

GLHF... If the 4.5 release ships with such severe bugs it won't be included in any distro but Fedora...

If you check the release notes for each major and minor release you will see that they all have fixes for ICEs

Oh and if it wasn't for Fedora / Redhat we wouldn't have nearly as much effort going into GCC and fixes going into packages to make them compile with the latest versions.

Try looking in the source RPMs / debs / ebuilds and see who provided the patches applied. I bet more than half of them are provided by Redhat

Fedora takes the hard road and helps stabilise packages before the other distros jump on the bandwagon. If they didn't do this things simply wouldn't be as stable over the whole Linux ecosystem

oh and I'm a Gentoo user not a Fedora fanboy