Date: Fri, 23 Aug 2019 11:26:07 -0700 From: Conrad Meyer <cem@freebsd.org> To: "Bjoern A. Zeeb" <bz@freebsd.org> Cc: Warner Losh <imp@bsdimp.com>, Li-Wen Hsu <lwhsu@freebsd.org>, John Baldwin <jhb@freebsd.org>, src-committers <src-committers@freebsd.org>, svn-src-all <svn-src-all@freebsd.org>, svn-src-head <svn-src-head@freebsd.org> Subject: Re: svn commit: r351364 - in head/sys: crypto/blowfish crypto/chacha20 crypto/des opencrypto Message-ID: <CAG6CVpVoJ_SLJy%2BMKB84fokuj89mfYDMLZJC7vZuK2efTyEXhw@mail.gmail.com> In-Reply-To: <D9BE08F9-7513-43E7-91D9-27814115BE43@FreeBSD.org> References: <201908220002.x7M028Jh070116@repo.freebsd.org> <DC49D260-C049-47F3-ADFA-76D793212E16@FreeBSD.org> <0b9d1aa1-d328-30bc-b939-f1407e236855@FreeBSD.org> <CANCZdfq7FOsQNkpRgHZP3yzxhfcGv5gLOzco7tGttTVSrETOfQ@mail.gmail.com> <3EE09B22-254B-4415-8865-D9542122ACA5@FreeBSD.org> <CANCZdfq6CRQcZuMr6X%2BTyFwCZm2FFx7qSbZcq%2BD4sUsbE7x3pg@mail.gmail.com> <D9BE08F9-7513-43E7-91D9-27814115BE43@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
At expected peril of wading into a thread >4 emails deep,
@Warner, modern GCC reports a similar warning; it just doesn't become
an error (at least in CI?). I'm not sure of the mechanism. Maybe
CI-specific? Old GCC didn't have a -Wno-error for -Wcast-qual, so
-Wcast-qual + -Werror there produced an error; that's why
$NO_WCAST_QUAL in conf/kern.mk is defined to -Wno-cast-qual for old
GCC but -Wno-error=3Dcast-qual for newer compilers. That said, this
file does not appear to be compiled with ${NO_WCAST_QUAL} either way.
@Bjoern,
So... why does GCC warn about this? key is const uint8_t*. The cast
is to const des_cblock *. I think the problem is that des_cblock is
defined as 'unsigned char [8]', so a 'const des cblock *' is actually
a 'const unsigned char**'? So... I think basically the entire des
subsystem may be accidentally using the wrong pointer level
throughout? The constify change just exposes that because correct
const-preserving cast of 'const foo*' to 'foo**' would be 'foo * const
*' (if I'm understanding this correctly).
Maybe one more reason to excise des from the tree.
Best,
Conrad
On Fri, Aug 23, 2019 at 9:59 AM Bjoern A. Zeeb <bz@freebsd.org> wrote:
>
> On 23 Aug 2019, at 16:48, Warner Losh wrote:
> > There's a lot of -Wno-error and -Wno-error=3DXXX sprinkled in our build
> > for
> > gcc 4.2.1 today, so we see the warnings but aren't stopped by them. My
> > changes take a big hammer and add a global -Wno-error to CFLAGS last
> > to
> > make this the behavior on gcc 4.2.1 platforms.
>
>
> Yes, but that didn=E2=80=99t answer my questions. It doesn=E2=80=99t hel=
p to try to
> avoid undefined C behaviour.
>
> That jenkins build seems to use the toolchain from ports and with that
> gcc 6.4.0.
>
> We see the same warning but it didn=E2=80=99t error as it seems to have d=
one
> for other architectures with the in-tree gcc with the same warnings:
>
> In file included from /workspace/src/sys/opencrypto/xform.c:94:0:
> /workspace/src/sys/opencrypto/xform_des1.c: In function 'des1_setkey':
> /workspace/src/sys/opencrypto/xform_des1.c:102:15: warning: cast
> discards 'const' qualifier from pointer target type [-Wcast-qual]
> des_set_key((const des_cblock *) key, p[0]);
> ^
> In file included from /workspace/src/sys/opencrypto/xform.c:95:0:
> /workspace/src/sys/opencrypto/xform_des3.c: In function 'des3_setkey':
> /workspace/src/sys/opencrypto/xform_des3.c:103:15: warning: cast
> discards 'const' qualifier from pointer target type [-Wcast-qual]
> des_set_key((const des_cblock *)(key + 0), p[0]);
> ^
> /workspace/src/sys/opencrypto/xform_des3.c:104:15: warning: cast
> discards 'const' qualifier from pointer target type [-Wcast-qual]
> des_set_key((const des_cblock *)(key + 8), p[1]);
> ^
> /workspace/src/sys/opencrypto/xform_des3.c:105:15: warning: cast
> discards 'const' qualifier from pointer target type [-Wcast-qual]
> des_set_key((const des_cblock *)(key + 16), p[2]);
> ^
> --
>
>
>
> To me this means that we treat different versions of compilers (in-tree
> and out-of-tree, gcc vs. clang) too different.
>
> If two versions of gcc (before your commit) gave the same warning I
> would have expected them to equally fail and not one fail and one pass?
>
> My question was: why was that the case?
>
> /bz
>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAG6CVpVoJ_SLJy%2BMKB84fokuj89mfYDMLZJC7vZuK2efTyEXhw>