Re: [PATCH 10/16] lib: General Linux RV64 syscall support.

[Jan Nieuwenhuizen]

W. J. van der Laan writes:

Hello,

> On Wednesday, May 5th, 2021 at 9:45 PM, Jan Nieuwenhuizen <janneke@gnu.org> 
> wrote:
>
>> > I think the decision what kernel to set as minimum depends on whether
>> > it is valuable to be able to bootstrap on really old kernels on
>> > i386. I have no insight into this.
>>
>> Yes, that's an interesting perspective. However, if we can add RISC-V
>> support without adding code duplication or extra parameterization, I
>> think that's a real benefit. We can always add support for older
>> Linux kernels when the need arises.
>>
>> I just chose the simple variants at the time because they were familiar
>> to me.
>
> So to be clear: you prefer that the newer syscalls, as introduced for
> RISC-V, are used for all platforms?

Yes, I think so.  Looking at your RISC-V patch, it suggested that we
would only have to remove all the #if branches for the old calls,
and (possibly most work) add the syscall defines for the new calls.

> Currently the i386/arm/etc syscall codepaths are completely
> unchanged. I expected that would be preferred to minimize the chance
> of breaking anything.

Yes, of course we will still need to check the other architectures;
before merging such a change I want to verify if the whole guix
bootstrap still works.  Otoh, it should "just work", right?

> I don't think variability between platforms with regard to syscalls
> can be completely avoided (there are some differences, especially
> between 32 bit and 64 bit), but it is definitely possible to bring
> that down as much as possible.

Let's see where this goes, the current situation works for us, I just
like it very much if we can avoid making the code base more complex than
necessary :-)

Greetings,
Janneke

-- 
Jan Nieuwenhuizen <janneke@gnu.org> | GNU LilyPond http://lilypond.org
Freelance IT http://JoyofSource.com | Avatar® http://AvatarAcademy.com