What is the best way to compare an unsigned and a constant?

From: SeongJae Park
Date: Fri Dec 27 2019 - 07:40:41 EST


Hello,


I have a function returning 'unsigned long', and would like to write a kunit
test for the function, as below.

unsigned long foo(void)
{
return 42;
}

static void foo_test(struct kunit *test)
{
KUNIT_EXPECT_EQ(test, 42, foo());
}

However, this kunit gives me below warning for the above code:

/.../linux/include/linux/kernel.h:842:29: warning: comparison of distinct pointer types lacks a cast
(!!(sizeof((typeof(x) *)1 == (typeof(y) *)1)))
^
/.../linux/include/kunit/test.h:493:9: note: in expansion of macro â__typecheckâ
((void)__typecheck(__left, __right)); \
^~~~~~~~~~~
/.../linux/include/kunit/test.h:517:2: note: in expansion of macro âKUNIT_BASE_BINARY_ASSERTIONâ
KUNIT_BASE_BINARY_ASSERTION(test, \
^~~~~~~~~~~~~~~~~~~~~~~~~~~
/.../linux/include/kunit/test.h:606:2: note: in expansion of macro âKUNIT_BASE_EQ_MSG_ASSERTIONâ
KUNIT_BASE_EQ_MSG_ASSERTION(test, \
^~~~~~~~~~~~~~~~~~~~~~~~~~~
/.../linux/include/kunit/test.h:616:2: note: in expansion of macro âKUNIT_BINARY_EQ_MSG_ASSERTIONâ
KUNIT_BINARY_EQ_MSG_ASSERTION(test, \
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/.../linux/include/kunit/test.h:979:2: note: in expansion of macro âKUNIT_BINARY_EQ_ASSERTIONâ
KUNIT_BINARY_EQ_ASSERTION(test, KUNIT_EXPECTATION, left, right)
^~~~~~~~~~~~~~~~~~~~~~~~~
/.../linux/mm/foo-test.h:565:2: note: in expansion of macro âKUNIT_EXPECT_EQâ
KUNIT_EXPECT_EQ(test, 42, foo());
^~~~~~~~~~~~~~~

I could remove the warning by explicitly type casting the constant as below:

KUNIT_EXPECT_EQ(test, (unsigned long)42, foo());

However, now 'checkpatch.pl' complains about the type casting as below.

WARNING: Unnecessary typecast of c90 int constant
#565: FILE: mm/foo-test.h:565:
+ KUNIT_EXPECT_EQ(test, (unsigned long)42, foo());

Of course, there could be several work-arounds for these warnings, such as
using 'EXPECT_TRUE(test, 42 == foo())' or casting the function's return value.
Nonetheless, I'm not sure what is the right way. Could you please let me know
what is the recommended way for this case?


Thanks,
SeongJae Park