Re: X86 GIT GCC 5 compilation warning

From: Michael Matz <matz@suse.de>
To: Borislav Petkov <bp@alien8.de>
Cc: Enrico Mioso <mrkiko.rs@gmail.com>,
	linux-kernel@vger.kernel.org, x86-ml <x86@kernel.org>
Subject: Re: X86 GIT GCC 5 compilation warning
Date: Tue, 16 Jun 2015 13:59:20 +0200 (CEST)	[thread overview]
Message-ID: <alpine.LSU.2.20.1506161353270.27315@wotan.suse.de> (raw)
In-Reply-To: <20150615171750.GP4255@pd.tnic>

Hello,

On Mon, 15 Jun 2015, Borislav Petkov wrote:

> Hmm, so I did start my oS13.2 i386 guest:
> 
> $ as --version
> GNU assembler (GNU Binutils; openSUSE 13.2) 2.24.0.20140403-6.1

That won't show the issue.  Our binutils are compiled with support for 
multiple targets, among them 64bit ones, and as I said, that makes the 
internal value types be 64bit as well, and hence masks the warning.

You could compile binutils yourself in that i386 guest, without enabling 
other than the host target to see the problem, but why would you want 
that?

> $ gcc --version

Again, gcc doesn't enter the picture here, it's gas itself that warns if 
it was compiled with the right (or wrong) options.

> $ cat t.s
> .text
> LOWMEM_PAGES = (((1<<32) - 0xc0000000) >> 12)
> 
>         mov LOWMEM_PAGES, %eax
> 
> $ as t.s
> $

My self-compiled gas warns here.

> Do we have some bleeding edge gcc5 rpms somewhere I could try?

devel:gcc, but gcc is not the cause here.

Ciao,
Michael.