2008-06-02 08:00:14

by Adrian Bunk

[permalink] [raw]
Subject: Re: initialization of static per-cpu variables

On Sun, May 25, 2008 at 08:35:24PM +1000, Rusty Russell wrote:
> On Saturday 24 May 2008 00:29:04 Adrian Bunk wrote:
> > On Thu, May 22, 2008 at 06:20:06PM +1000, Rusty Russell wrote:
> > > Yep, it was an old toolchain used by Sparc: DaveM found this one. As you
> > > say, it's ancient: I'm happy to queue a cleanup patch now everyone is on
> > > a modern compiler.
> >
> > The commit says:
> >
> > GCC3.1 apparently gets confused about uninitialized sections
> >
> > We do still support gcc 3.2 (which is the same as 3.1 except for a C++
> > ABI change) as a compiler for the kernel.
>
> Adrian, that's a little silly. There are obviously bug fixes in 3.2 over
> 3.1.0.

I've checked the announcements of 3.1.1 and 3.2, and at least for me
nothing looked like it would fix this bug.

> Noone has complained about the introduction of multiple other cases
> which would screw things up if they experienced this bug.

I doubt there is any serious userbase for gcc 3.2 left.

But your "now everyone is on a modern compiler" does not match what we
announce as supported compiler versions for the kernel.

If you have a good reason for pushing the minimum required gcc version
for compiling the kernel to 3.3 or 3.4 [1] you have my full support, but
as long as we officially support gcc 3.2 we should try to break as few
as possible - especially since it will take time until anyone will run
into any breakage.

> Finally, it's a sparc64 problem and DaveM acked already.

Was it a sparc64 problem or a generic problem DaveM happened to run into
on sparc?

> That's half the userbase!

sparc64 isn't that unpopular...

> Cheers,
> Rusty.

Sorry if I'm sounding overly pedantic, but I want that what we try as
good as possible that what we announce as being supported also works
(even if this results in several workarounds shipped).

cu
Adrian

[1] gcc 3.4 still has a serious userbase at least in ARM country, so you
won't be able to drop support for it

--

"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed


2008-06-02 23:11:48

by Rusty Russell

[permalink] [raw]
Subject: Re: initialization of static per-cpu variables

On Monday 02 June 2008 17:59:37 Adrian Bunk wrote:
> On Sun, May 25, 2008 at 08:35:24PM +1000, Rusty Russell wrote:
> > Finally, it's a sparc64 problem and DaveM acked already.
>
> Was it a sparc64 problem or a generic problem DaveM happened to run into
> on sparc?

Yes, it was a sparc specific problem.

Rusty.