linux-kernel - __diag_ignore_all(), GCC < 8, extra warnings, and -Werror

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <87wmzezns4.fsf@intel.com>
Date:   Wed, 05 Jul 2023 12:54:35 +0300
From:   Jani Nikula <jani.nikula@...el.com>
To:     linux-kernel@...r.kernel.org, workflows@...r.kernel.org,
        linux-kbuild@...r.kernel.org, intel-gfx@...ts.freedesktop.org
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        Rodrigo Vivi <rodrigo.vivi@...el.com>,
        Tvrtko Ursulin <tvrtko.ursulin@...ux.intel.com>,
        Joonas Lahtinen <joonas.lahtinen@...ux.intel.com>,
        Masahiro Yamada <masahiroy@...nel.org>
Subject: __diag_ignore_all(), GCC < 8, extra warnings, and -Werror

For a long time now, i915 has enabled a bunch of W=1 style warnings
locally, and we try hard to keep i915 warning free.

One of the warnings is -Woverride-init from -Wextra. We need to bypass
that in a few cases, and used to do this for the relevant files:

	CFLAGS_file.o = $(call cc-disable-warning, override-init)

Recently, we switched from the above to a more localized version in each
file.c:

	__diag_push();
	__diag_ignore_all("-Woverride-init", "Allow overriding inherited members");
	...
        __diag_pop();

We now got a report that this fails the build with CONFIG_WERROR=y or
W=e when using GCC version < 8. Indeed, __diag_ignore_all() requires GCC
version 8 or later.

Should we now revert back to disabling -Woverride-init on a file
granularity? Should we consider breaking the build for CONFIG_WERROR=y
or W=e on older compilers a regression?

I'll note that with the current usage of __diag_ignore_all() elsewhere
in kernel, CONFIG_WERROR=y or W=e with W=1 will never pass on older
compilers. But then again, it has never passed on any compiler, so it
can't be a regression.

BR,
Jani.

-- 
Jani Nikula, Intel Open Source Graphics Center