GNU's Gold Linker Is Stagnating, Fedora Looking To Punt It Off Into A Separate Package

[phoronix]

Phoronix: GNU's Gold Linker Is Stagnating, Fedora Looking To Punt It Off Into A Separate Package

While the GNU Gold linker has been quite promising especially in being faster than the conventional GNU linker, Google developers are no longer actively advancing this linker and thus raising concerns it could begin to suffer from bit-rot...

http://www.phoronix.com/scan.php?page=news_item&px=GNU-Gold-Stagnate-F31

[RealNC]

Is the normal linker marching forward? Michael, when is the last time you posted news about the normal linker getting advanced?

If the normal linker isn't being "advanced" either, then drop both linkers. We won't be able to link anymore but at least we get rid of linkers that are not "advancing."

[WesternSemiconductor]

Is the normal linker marching forward?

Yes.

It also happens to be maintained by the same group of people maintaining binutils. Right now gold is maintained by, effectively, nobody. When interfaces it depends on evolve somebody needs to make those updates, or else it breaks.

Michael, when is the last time you posted news about the normal linker getting advanced?

I don't think mundane day-to-day maintenance is really newsworthy, do you? At least not when it's happening. It's like toilets flushing properly, nobody complains when it's happening.

The real issue here is Google's attitude that it's okay to abandon projects whenever it gets bored. Cough, cough, Google Plus, Cough. That attitude doesn't fly when it comes to compiler infrastructure, which has multi-decade lifespans. If you can't fund the maintenance then you can't really afford that gee-whiz rewrite-from-scratch either.

[xen0n]

Just another prime manifestation of Google's attitude towards GPLed software. This has happened multiple times, especially in AOSP where every single GPL component got rewritten or replaced (BlueZ -> bluedroid IIRC, busybox -> toybox, gcc -> clang, and ultimately Linux -> Fuchsia I believe).

Simply WTF.

[Gusar]

Just another prime manifestation of Google's attitude towards GPLed software. This has happened multiple times, especially in AOSP where every single GPL component got rewritten or replaced (BlueZ -> bluedroid IIRC, busybox -> toybox, gcc -> clang, and ultimately Linux -> Fuchsia I believe).

Android has had a "no GPL in userspace" policy since the very beginning. They never used busybox precisely for that reason, they had a NIH box called toolbox before switching to toybox. They used BlueZ only reluctantly, because there was no other choice. So yeah, Google isn't favorable to GPL to put it mildly, but regarding Android that's nothing new.

Google dropping projects whenever they feel like it is also nothing new. It's also quite difficult to reuse their stuff in your own project, because Google is big on static linking (a big no-no for Linux distributions), doesn't care about API/ABI stability, doesn't know about soname versioning (why would they when they just link everything statically), etc.

[brad0]

Just another prime manifestation of Google's attitude towards GPLed software. This has happened multiple times, especially in AOSP where every single GPL component got rewritten or replaced (BlueZ -> bluedroid IIRC, busybox -> toybox, gcc -> clang, and ultimately Linux -> Fuchsia I believe).

Simply WTF.

The concept is so simple it went right over your head. Wooosh. And it's far from just Google.

[brad0]

The real issue here is Google's attitude that it's okay to abandon projects whenever it gets bored. Cough, cough, Google Plus, Cough. That attitude doesn't fly when it comes to compiler infrastructure, which has multi-decade lifespans. If you can't fund the maintenance then you can't really afford that gee-whiz rewrite-from-scratch either.

They have moved on to better options. It flies just fine. It doesn't bother them when they have moved on to better options. They have nothing to worry about and they have no obligation to do anything.

[brent]

Google dropping contributions to OSS software they no longer need is a normal business decision. I don't see anything wrong with it. They have no obligations to continue working on something. Google isn't special here either, all sane companies behave like that.

In the same way, they are free to prefer permissive licenses as opposed to GPL. There is nothing wrong with it at all. Sorry, but the software world doesn't revolve around the GPL. There are all kinds of licensing models, and they can coexist just fine.

[smitty3268]

Why is the GNU linker still active, rather than everyone moving on towards gold?

I haven't looked into it at all so I could be missing something, but it seems to be both more performant, and you'd think designed much more cleanly (and recently) and therefore more maintainable.

Is there something in particular it still can't do at all and would be difficult to implement? Is it just NIH by people who've been working on the old linker for 20 years? A fear of any kind of change away from the status quo?

[chithanh]

Why is the GNU linker still active, rather than everyone moving on towards gold?

Because lots of software is broken when built with the gold linker.

https://bugs.gentoo.org/269315