[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <a55cb545-30d8-c035-ec23-4e94c132d7ba@suse.cz>
Date: Mon, 17 Oct 2022 14:23:34 +0200
From: Martin Liška <mliska@...e.cz>
To: David Laight <David.Laight@...LAB.COM>,
'Andrew Morton' <akpm@...ux-foundation.org>,
Peter Oberparleiter <oberpar@...ux.ibm.com>
Cc: "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] gcov: support GCC 12.1 and newer compilers
On 10/17/22 13:38, David Laight wrote:
> From: Andrew Morton
>> Sent: 17 October 2022 01:41
> ..
>> The changelog doesn't tell us what the user-visible effects of this are
>> (please, it should do so), but it sounds to me like those effects are
>> "gcov is utterly busted".
>>
>> So I'll add a cc:stable to this, so that people can use new gcc
>> versions to build older kernels.
>
> I can't help wondering what happens if you link a binary
> compiled with an old gcc with one build with a new gcc?
It should be fine as the code emits each .gcda file with version that
comes from the compilation phase:
pos += store_gcov_u32(buffer, pos, info->version);
>
> This could easily happen with out of tree loadable modules.
Sure.
> Or just linking an old .a file into a userspace binary.
>
> Now maybe the gcov data isn't used (I've not looked up what
> it is for) but is sounds like something whose format should
> be set in stone?
No, I've doing some changes to the format based on new challenges we face.
Cheers,
Martin
>
> David
>
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)
Powered by blists - more mailing lists