This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: PATCH: Add --size-check=[error|warning]
- From: Ingo Molnar <mingo at elte dot hu>
- To: Jan Beulich <JBeulich at novell dot com>
- Cc: sedat dot dilek at gmail dot com, Alan Modra <amodra at bigpond dot net dot au>, "H.J. Lu" <hjl dot tools at gmail dot com>, Thomas Gleixner <tglx at linutronix dot de>, Andrew Morton <akpm at linux-foundation dot org>, Linus Torvalds <torvalds at linux-foundation dot org>, binutils <binutils at sourceware dot org>, LKML <linux-kernel at vger dot kernel dot org>, "H. Peter Anvin" <hpa at zytor dot com>
- Date: Mon, 14 Mar 2011 12:35:24 +0100
- Subject: Re: PATCH: Add --size-check=[error|warning]
- References: <AANLkTinDTyupNgZR-0izCWrpwOnE1wEcubn+M3o=Mgy2@mail.gmail.com> <4D7E095802000078000363D5@vpn.id2.novell.com>
* Jan Beulich <JBeulich@novell.com> wrote:
> >>> On 14.03.11 at 12:02, Sedat Dilek <sedat.dilek@googlemail.com> wrote:
> > [1] Jan, what do you mean by "side-effects". Can you explain that a
> > bit more precisely?
>
> The -B compiler option controls more than just finding "helper" binaries.
>
> > [2] Where can someone set a "global behaviour" (hardcoded options) for
> > his/her assembler in the kernel's build-system (speaking of
> > "--size-check=[error|warning]")?
>
> Nowhere, selecting behavior is possible only via the command line.
>
> > [3] Can the kernel-buildsystem check for system's binutils/as version
> > and/or its features/options? If yes, where would that be and can you
> > offer a snippet for a solution?
>
> Making the kernel build system check for certain newly introduced
> gas options would again require changes to the kernel sources,
> which is precisely what is impossible to do for past kernel releases
> (and bisection in particular).
Yes, and all the counter-arguments here continue to miss that very simple point.
That point was made in the first post about this topic and it's still not
acknowledged - let alone addressed.
This breakage is unnecessary and retroactively goes back 130,000 commits. A warning
combined with not issuing the debug symbol would be just as fine and would still
result in valid output.
Thanks,
Ingo