Re: [PATCH 2/3] kunit: add KUnit array assertions to the example_all_expect_macros_test

From: Daniel Latypov <dlatypov@google.com>
To: "André Almeida" <andrealmeid@riseup.net>
Cc: "Maíra Canal" <mairacanal@riseup.net>,
	melissa.srw@gmail.com, daniel@ffwll.ch, javierm@redhat.com,
	siqueirajordao@riseup.net,
	"Isabella Basso" <isabbasso@riseup.net>,
	jose.exposito89@gmail.com, magalilemes00@gmail.com,
	tales.aparecida@gmail.com, davidgow@google.com,
	davem@davemloft.net,
	"Brendan Higgins" <brendanhiggins@google.com>,
	linux-kselftest@vger.kernel.org, kunit-dev@googlegroups.com,
	linux-kernel@vger.kernel.org, airlied@linux.ie, kuba@kernel.org
Subject: Re: [PATCH 2/3] kunit: add KUnit array assertions to the example_all_expect_macros_test
Date: Tue, 2 Aug 2022 11:15:30 -0700	[thread overview]
Message-ID: <CAGS_qxodPndQZ_ypy-QP=ViNUvwZk1z1u8EAv9k5XzDEC4WSGQ@mail.gmail.com> (raw)
In-Reply-To: <5c9038fd-a247-d0d3-841c-ba1e606bc309@riseup.net>

On Tue, Aug 2, 2022 at 9:19 AM André Almeida <andrealmeid@riseup.net> wrote:
> Às 13:12 de 02/08/22, Maíra Canal escreveu:
> > Increament the example_all_expect_macros_test with the
> > KUNIT_EXPECT_ARREQ and KUNIT_EXPECT_ARRNEQ macros by creating a test
> > with array assertions.
> >
> > Signed-off-by: Maíra Canal <mairacanal@riseup.net>
> > ---
> >  lib/kunit/kunit-example-test.c | 7 +++++++
> >  1 file changed, 7 insertions(+)
> >
> > diff --git a/lib/kunit/kunit-example-test.c b/lib/kunit/kunit-example-test.c
> > index f8fe582c9e36..fc81a45d9cbc 100644
> > --- a/lib/kunit/kunit-example-test.c
> > +++ b/lib/kunit/kunit-example-test.c
> > @@ -86,6 +86,9 @@ static void example_mark_skipped_test(struct kunit *test)
> >   */
> >  static void example_all_expect_macros_test(struct kunit *test)
> >  {
> > +     const u32 array[] = { 0x0F, 0xFF };
> > +     const u32 expected[] = { 0x1F, 0xFF };

Given the distance between the definition and their use, perhaps we
can give them clearer names.
E.g. array + diff_array, or array1 + array2, etc.

I think something to indicate they're arrays and that they're different.
The current name `expected` is a bit unclear.

> > +
> >       /* Boolean assertions */
> >       KUNIT_EXPECT_TRUE(test, true);
> >       KUNIT_EXPECT_FALSE(test, false);
> > @@ -109,6 +112,10 @@ static void example_all_expect_macros_test(struct kunit *test)
> >       KUNIT_EXPECT_STREQ(test, "hi", "hi");
> >       KUNIT_EXPECT_STRNEQ(test, "hi", "bye");
> >
> > +     /* Array assertions */
> > +     KUNIT_EXPECT_ARREQ(test, expected, expected, 2);
> > +     KUNIT_EXPECT_ARRNEQ(test, array, expected, 2);
>
> ARRAY_SIZE() is usually better than constants is this case.

Note: that's actually incorrect!

Ah right, this was the other blocker I had in mind.
I wasn't sure how we'd handle the size parameter.

Users might think ARRAY_SIZE() is fine and copy-paste it.
But the size parameter is in units of bytes, not array elements!
If the element types are not 1 byte, it'll silently not compare the full array.

We'd want people to use
KUNIT_EXPECT_ARREQ(test, expected, expected, sizeof(expected));

But this doesn't work for `u32 *array`, since it'll silently just
compare 1 byte if people get them mixed up.

I don't know how we make a maximally fool-proof version of this macro :\