Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Message-ID: <1529353683.3092.32.camel@sipsolutions.net>
Subject: Re: [PATCH v3] bitfield: fix *_encode_bits()
From:   Johannes Berg <johannes@sipsolutions.net>
To:     Andy Shevchenko <andy.shevchenko@gmail.com>
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        netdev <netdev@vger.kernel.org>,
        Al Viro <viro@zeniv.linux.org.uk>
Date:   Mon, 18 Jun 2018 22:28:03 +0200
In-Reply-To: <CAHp75VdY+PF0Edm_py+UiD3nz77HTQps_8uaRh+Sa3C+UKECKA@mail.gmail.com> (sfid-20180618_222109_796813_69F8BCFA)
References: <20180618195618.17536-1-johannes@sipsolutions.net>
         <CAHp75VdY+PF0Edm_py+UiD3nz77HTQps_8uaRh+Sa3C+UKECKA@mail.gmail.com>
         (sfid-20180618_222109_796813_69F8BCFA)
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


> I think would be better to add test cases first, followed by fix. (1
> patch -> 2 patches)
> In this case Fixes tag would be only for the fix part and backporting
> (if needed) will be much easier.

Can't, unless I introduce a compilation issue in the tests first? That
seems weird. But I guess I can do it the other way around.

> > @@ -143,6 +143,7 @@ static __always_inline base type##_get_bits(__##type v, base field) \
> >         ____MAKE_OP(le##size,u##size,cpu_to_le##size,le##size##_to_cpu) \
> >         ____MAKE_OP(be##size,u##size,cpu_to_be##size,be##size##_to_cpu) \
> >         ____MAKE_OP(u##size,u##size,,)
> > +____MAKE_OP(u8,u8,,)
> 
> Is this one you need, or it's just for sake of tests?

All three ;-)

We'll probably need it eventually (we do have bytes to take bits out
of), for consistency I think we want it, and I wanted to add it to the
tests too.

> For me looks like for consistency we may add fake conversion macros
> for this, such as
> 
> #define cpu_to_le8(x) x
> #define le8_to_cpu(x) x
> ...
> #undef le8_to_cpu
> #undef cpu_to_le8
> 
> And do in the same way like below
> 
> __MAKE_OP(8)

I disagree with this. I don't see why we should have le8_encode_bits()
and be8_encode_bits() and friends, that makes no sense.

> Perhaps
> // SPDX... GPL-2.0+

Yeah, I guess I should have that.

> > +/*
> > + * Test cases for bitfield helpers.
> > + */
> > +
> > +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> > +
> > +#include <linux/init.h>
> > +#include <linux/kernel.h>
> > +#include <linux/module.h>
> 
> Either module.h (if we can compile as a module) or just init.h otherwise.

It can be a module ... guess I cargo-culted that from another test.

> > +/*
> > + * This should fail compilation:
> > + *     CHECK_ENC_GET(16, 16, 0x0f00, 0x1000);
> > + */
> 
> Perhaps we need some ifdeffery around this. It would allow you to try
> w/o altering the source code.
> 
> #ifdef TEST_BITFIELD_COMPILE
> ...
> #endif

Yeah, I guess we could do that.

> I guess you rather continue and print a statistics X passed out of Y.
> Check how it's done, for example, in other test_* modules.
> (test_printf.c comes first to my mind).

I see it's done that way elsewhere, but I don't really see the point. It
makes the test code more complex, and if you fail here you'd better fix
it, and if you need a few iterations for that it's not really a problem?

johannes