Re: RFC: Raise required gcc version to 3.2 ?

Re: RFC: Raise required gcc version to 3.2 ?

[Mikael Pettersson]

On Thu, 28 Jul 2005 14:00:12 +0200, Adrian Bunk wrote:
>What is the oldest gcc we want to support in kernel 2.6?
>
>Currently, it's 2.95 .
>
>I'd suggest raising this to 3.2 which should AFAIK not be a problem for 
>any distribution supporting kernel 2.6 .
>
>Is there any good reason why we should not drop support for older 
>compilers?

You're asking the wrong question. The right question would be:
"Is there any good reason to drop support for older compilers?"

At least on i386, gcc-2.95.3 still works and has the advantage
of being much faster compile-time wise on older machines with
limited memory (like my 486 test box). And I'm not the only
one with that POV -- akpm also uses 2.95.

Of course, if keeping 2.95.3 support would somehow hinder the
kernel's development, then it should be removed. But so far I
haven't seen any real(*) evidence that this is the case.

(*) Moronic code with declarations after statements is not a
valid argument against 2.95.

/Mikael

Re: RFC: Raise required gcc version to 3.2 ?

[Adrian Bunk]

On Thu, Jul 28, 2005 at 06:48:49PM +0200, Mikael Pettersson wrote:
> On Thu, 28 Jul 2005 14:00:12 +0200, Adrian Bunk wrote:
> >What is the oldest gcc we want to support in kernel 2.6?
> >
> >Currently, it's 2.95 .
> >
> >I'd suggest raising this to 3.2 which should AFAIK not be a problem for 
> >any distribution supporting kernel 2.6 .
> >
> >Is there any good reason why we should not drop support for older 
> >compilers?
> 
> You're asking the wrong question. The right question would be:
> "Is there any good reason to drop support for older compilers?"
> 
> At least on i386, gcc-2.95.3 still works and has the advantage
> of being much faster compile-time wise on older machines with
> limited memory (like my 486 test box). And I'm not the only
> one with that POV -- akpm also uses 2.95.
> 
> Of course, if keeping 2.95.3 support would somehow hinder the
> kernel's development, then it should be removed. But so far I
> haven't seen any real(*) evidence that this is the case.
>...

The advantages are:
- reducing the number of supported gcc versions from 8 to 4 [1]
  allows the removal of several #ifdef's and workarounds
- my impression is that the older compilers are only rarely
  used, so miscompilations of a driver with an old gcc might
  not be detected for a longer amount of time

My personal opinion about the time and space a compilation requires is 
that this is no longer that much of a problem for modern hardware, and 
in the worst case you can compile the kernels for older machines on more 
recent machines.

> /Mikael

cu
Adrian

[1] support removed: 2.95, 2.96, 3.0, 3.1
    still supported: 3.2, 3.3, 3.4, 4.0

-- 

       "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

RFC: Raise required gcc version to 3.2 ?

[Adrian Bunk]

What is the oldest gcc we want to support in kernel 2.6?

Currently, it's 2.95 .

I'd suggest raising this to 3.2 which should AFAIK not be a problem for 
any distribution supporting kernel 2.6 .

Is there any good reason why we should not drop support for older 
compilers?

cu
Adrian

-- 

       "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

Re: RFC: Raise required gcc version to 3.2 ?

[Krzysztof Halasa]

Adrian Bunk <bunk@stusta.de> writes:

> Is there any good reason why we should not drop support for older 
> compilers?

Compilation speed? Don't know, using 3 (4?) years old Athlon 2000
it's not a problem unless I need full build 30 times a day.

But people on 266 MHz ARM5 may notice.
-- 
Krzysztof Halasa

Re: RFC: Raise required gcc version to 3.2 ?

[Lennart Sorensen]

On Thu, Jul 28, 2005 at 02:39:12PM +0200, Krzysztof Halasa wrote:
> Compilation speed? Don't know, using 3 (4?) years old Athlon 2000
> it's not a problem unless I need full build 30 times a day.
> 
> But people on 266 MHz ARM5 may notice.

Hmm, an a 400Mhz PXA 255 I found using gcc 3.x meant I could use xscale
rather than older optimizations, and the resulting kernel certainly felt
much faster (I never did actually meassure if it was faster or not).
Using gcc 2.95 requried changing the cpu optimization lines in the
kernel.  Of course this was in a kernel patched with the pxa arm patches
so it isn't quite the same as the mainline kernel.  I certainly won't
consider using gcc 2.95 on an arm anymore (although perhaps on older arm
models it does make sense).

I have no idea if it took longer to compile with gcc 3.x, but the result
seemed better to me.  To me performance of the resulting kernel is more
important than performance compiling the kernel.

Of course if you leave the choice as 2.95 and above rather than 3.2 and
above, people can pick whatever they want based on their needs.

Len Sorensen

Re: RFC: Raise required gcc version to 3.2 ?

[Christoph Lameter]

On Thu, 28 Jul 2005, Adrian Bunk wrote:

> What is the oldest gcc we want to support in kernel 2.6?
> 
> Currently, it's 2.95 .
> 
> I'd suggest raising this to 3.2 which should AFAIK not be a problem for 
> any distribution supporting kernel 2.6 .

You have all my support for this. Some weird macros and code 
could be removed from the tree. F.e. One could use [] instead of [0] and 
there is a baaad macro in include/linux/mmzone.h in need of some healing 
touches.

> Is there any good reason why we should not drop support for older 
> compilers?

Probably but we should drop support anyways.