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

From: "Maíra Canal" <mairacanal@riseup.net>
To: "Daniel Latypov" <dlatypov@google.com>,
	"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 16:00:31 -0300	[thread overview]
Message-ID: <5931f10d-09ac-84f3-f5d1-65039478ae97@riseup.net> (raw)
In-Reply-To: <CAGS_qxodPndQZ_ypy-QP=ViNUvwZk1z1u8EAv9k5XzDEC4WSGQ@mail.gmail.com>

On 8/2/22 15:15, 'Daniel Latypov' via KUnit Development wrote:
> 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.

Thank you for the note, I'll address it at v2.

> 
>>> +
>>>       /* 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!
> 

Yep, that's my bad!

> 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 :\

This is a hard one also. I believe that use KUNIT_EXPECT_ARREQ(test,
expected, expected, sizeof(expected)); is more compliant to the
memcpy/memset/memcmp signature. Moreover, this problem also occur for
the KUNIT_EXPECT_EQ(test, memcmp(expected, expected, sizeof(expected)), 0);

I believe that the number of array elements will make it easier for
users to avoid mistakes.

I'll change it internally for size_bytes = (size) * sizeof((left)[0]) on v2.

Best Regards,
- Maíra Canal

> 