Use of `gawk -M' in kernel build scripts

Use of `gawk -M' in kernel build scripts

[Aharon Robbins]

Hi All

I'm the maintainer for gawk. I see that it's used in the kernel build
scripts in combination with objdump.

In particular, I note that gawk is invoked with the -M option, which
uses MPFR for arithmetic (although the precision isn't changed beyond
the default which is the same as for IEEE hardware floating point).

I am wondering if the -M option is really needed here?  The reason I
ask is that I plan to deprecate MPFR support. The upcoming 6.0 release
(I hope in Q2 2027) will issue a deprecation warning when -M is used,
and 6.1 (late 2027 or early 2028, I hope) will have MPFR support removed
entirely.

So, I wanted to give you folks a heads-up. If -M isn't needed, now's a good
time to remove its use from the kernel build.

Please reply privately, as I'm not on the linux-kbuild list.

Much thanks,

Arnold Robbins

Re: Use of `gawk -M' in kernel build scripts

[Nathan Chancellor]

Hi Aharon,

On Sun, Oct 04, 2026 at 09:26:58PM +0300, Aharon Robbins wrote:
> I'm the maintainer for gawk. I see that it's used in the kernel build
> scripts in combination with objdump.

For the record, the only place where I see '-M' used with gawk is in
scripts/objdump-func, which is not called during the kernel build
thankfully.

> In particular, I note that gawk is invoked with the -M option, which
> uses MPFR for arithmetic (although the precision isn't changed beyond
> the default which is the same as for IEEE hardware floating point).
> 
> I am wondering if the -M option is really needed here?  The reason I
> ask is that I plan to deprecate MPFR support. The upcoming 6.0 release
> (I hope in Q2 2027) will issue a deprecation warning when -M is used,
> and 6.1 (late 2027 or early 2028, I hope) will have MPFR support removed
> entirely.
> 
> So, I wanted to give you folks a heads-up. If -M isn't needed, now's a good
> time to remove its use from the kernel build.

Adding Josh and Peter for the rest of this since they have
added/supported this script.

-- 
Cheers,
Nathan

Re: Use of `gawk -M' in kernel build scripts

[arnold]

Hi Nathan.

Thanks for the note. gawk -M is also used in a file named check.c (I didn't
save the full path, sorry), which has the objdump-func script embedded
in it for calling via system(3).

Josh and Peter, looking forward to hearing more.

Thanks,

Arnold

Nathan Chancellor <nathan@kernel.org> wrote:

> Hi Aharon,
>
> On Sun, Oct 04, 2026 at 09:26:58PM +0300, Aharon Robbins wrote:
> > I'm the maintainer for gawk. I see that it's used in the kernel build
> > scripts in combination with objdump.
>
> For the record, the only place where I see '-M' used with gawk is in
> scripts/objdump-func, which is not called during the kernel build
> thankfully.
>
> > In particular, I note that gawk is invoked with the -M option, which
> > uses MPFR for arithmetic (although the precision isn't changed beyond
> > the default which is the same as for IEEE hardware floating point).
> > 
> > I am wondering if the -M option is really needed here?  The reason I
> > ask is that I plan to deprecate MPFR support. The upcoming 6.0 release
> > (I hope in Q2 2027) will issue a deprecation warning when -M is used,
> > and 6.1 (late 2027 or early 2028, I hope) will have MPFR support removed
> > entirely.
> > 
> > So, I wanted to give you folks a heads-up. If -M isn't needed, now's a good
> > time to remove its use from the kernel build.
>
> Adding Josh and Peter for the rest of this since they have
> added/supported this script.
>
> -- 
> Cheers,
> Nathan

Re: Use of `gawk -M' in kernel build scripts

[arnold]

Josh, Peter,

Anything to add?

Thanks,

Arnold

arnold@skeeve.com wrote:

> Hi Nathan.
>
> Thanks for the note. gawk -M is also used in a file named check.c (I didn't
> save the full path, sorry), which has the objdump-func script embedded
> in it for calling via system(3).
>
> Josh and Peter, looking forward to hearing more.
>
> Thanks,
>
> Arnold
>
> Nathan Chancellor <nathan@kernel.org> wrote:
>
> > Hi Aharon,
> >
> > On Sun, Oct 04, 2026 at 09:26:58PM +0300, Aharon Robbins wrote:
> > > I'm the maintainer for gawk. I see that it's used in the kernel build
> > > scripts in combination with objdump.
> >
> > For the record, the only place where I see '-M' used with gawk is in
> > scripts/objdump-func, which is not called during the kernel build
> > thankfully.
> >
> > > In particular, I note that gawk is invoked with the -M option, which
> > > uses MPFR for arithmetic (although the precision isn't changed beyond
> > > the default which is the same as for IEEE hardware floating point).
> > > 
> > > I am wondering if the -M option is really needed here?  The reason I
> > > ask is that I plan to deprecate MPFR support. The upcoming 6.0 release
> > > (I hope in Q2 2027) will issue a deprecation warning when -M is used,
> > > and 6.1 (late 2027 or early 2028, I hope) will have MPFR support removed
> > > entirely.
> > > 
> > > So, I wanted to give you folks a heads-up. If -M isn't needed, now's a good
> > > time to remove its use from the kernel build.
> >
> > Adding Josh and Peter for the rest of this since they have
> > added/supported this script.
> >
> > -- 
> > Cheers,
> > Nathan

Re: Use of `gawk -M' in kernel build scripts

[Peter Zijlstra]

On Mon, Oct 05, 2026 at 10:53:32AM +0200, Nathan Chancellor wrote:
> Hi Aharon,
> 
> On Sun, Oct 04, 2026 at 09:26:58PM +0300, Aharon Robbins wrote:
> > I'm the maintainer for gawk. I see that it's used in the kernel build
> > scripts in combination with objdump.
> 
> For the record, the only place where I see '-M' used with gawk is in
> scripts/objdump-func, which is not called during the kernel build
> thankfully.
> 
> > In particular, I note that gawk is invoked with the -M option, which
> > uses MPFR for arithmetic (although the precision isn't changed beyond
> > the default which is the same as for IEEE hardware floating point).
> > 
> > I am wondering if the -M option is really needed here?  The reason I
> > ask is that I plan to deprecate MPFR support. The upcoming 6.0 release
> > (I hope in Q2 2027) will issue a deprecation warning when -M is used,
> > and 6.1 (late 2027 or early 2028, I hope) will have MPFR support removed
> > entirely.
> > 
> > So, I wanted to give you folks a heads-up. If -M isn't needed, now's a good
> > time to remove its use from the kernel build.
> 
> Adding Josh and Peter for the rest of this since they have
> added/supported this script.

I'm not sure; Josh added that when he did the upstream version. My
initial version very much does not have that -M on. Josh?

Re: Use of `gawk -M' in kernel build scripts

[arnold]

Peter Zijlstra <peterz@infradead.org> wrote:

> On Mon, Oct 05, 2026 at 10:53:32AM +0200, Nathan Chancellor wrote:
> > Hi Aharon,
> > 
> > On Sun, Oct 04, 2026 at 09:26:58PM +0300, Aharon Robbins wrote:
> > > I'm the maintainer for gawk. I see that it's used in the kernel build
> > > scripts in combination with objdump.
> > 
> > For the record, the only place where I see '-M' used with gawk is in
> > scripts/objdump-func, which is not called during the kernel build
> > thankfully.
> > 
> > > In particular, I note that gawk is invoked with the -M option, which
> > > uses MPFR for arithmetic (although the precision isn't changed beyond
> > > the default which is the same as for IEEE hardware floating point).
> > > 
> > > I am wondering if the -M option is really needed here?  The reason I
> > > ask is that I plan to deprecate MPFR support. The upcoming 6.0 release
> > > (I hope in Q2 2027) will issue a deprecation warning when -M is used,
> > > and 6.1 (late 2027 or early 2028, I hope) will have MPFR support removed
> > > entirely.
> > > 
> > > So, I wanted to give you folks a heads-up. If -M isn't needed, now's a good
> > > time to remove its use from the kernel build.
> > 
> > Adding Josh and Peter for the rest of this since they have
> > added/supported this script.
>
> I'm not sure; Josh added that when he did the upstream version. My
> initial version very much does not have that -M on. Josh?

Thanks Peter.  Josh, looking forward to hearing from you.

Arnold