From mboxrd@z Thu Jan 1 00:00:00 1970
From: bugzilla-daemon@bugzilla.kernel.org
Subject: [Bug 207959] Don't warn about the universal zero initializer for a
structure with the 'designated_init' attribute.
Date: Thu, 28 May 2020 21:24:35 +0000
Message-ID:
References:
Mime-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8BIT
Return-path:
Received: from mail.kernel.org ([198.145.29.99]:40230 "EHLO mail.kernel.org"
rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
id S2436505AbgE1VYg (ORCPT );
Thu, 28 May 2020 17:24:36 -0400
In-Reply-To:
Sender: linux-sparse-owner@vger.kernel.org
List-Id: linux-sparse@vger.kernel.org
To: linux-sparse@vger.kernel.org
https://bugzilla.kernel.org/show_bug.cgi?id=207959
--- Comment #3 from Luc Van Oostenryck (luc.vanoostenryck@gmail.com) ---
(In reply to Asher Gordon from comment #2)
> Perhaps '-Wno-universal-initializer' should be the default?
Well, that's really the point.
The problem Sparse also gives the warnings corresponding to clang's -Wnonnull
and my understanding is that these warnings are desired for the kernel even
when coming from using '{ 0 }'.
> > My very personal point of view is that the correct syntax should be '{ }'
> > because it conveys much better the idea of a default initializer. This
> > single zero in '{ 0 }' is just confusing.
>
> I can see your point, but unfortunately, as Ramsay Jones says here[1] and
> Alexander Monakov here[2], this is not standard C. So '{ }' isn't an option
> if we want to be portable.
Yes, I know, it's a pity. It's why I said 'should be'.
> Andrew Pinski's suggestion[3] is also an option,
> but that seems ugly to me.
Yes, it's far from ideal.
> I'm writing a library, Mu[4], which has a structure for which the
> 'designated_init' attribute is appropriate (see the 'MU_OPT' structure
> here[5]). However, I don't want to force my users not to use '{ 0 }', which
> is why I think this feature would be useful.
Interesting.
Yes, I understand. Git was in the same kind of situation, it's why I added
'-Wno-universal-initializer'. Can't you just add this option in your
SPARSE_FLAGS or something like that?
> Also, a minor note: In the test program I attached, the attribute needs to
> be specified after the closing brace to work with Sparse.
Yes, it's a known problem. Sparse accept 'type attributes' (those situated just
after the keyword 'struct', 'union' or 'enum') but ignore them.
I've some unfinished patches for this ... since some time already :(
--
You are receiving this mail because:
You are watching the assignee of the bug.