tests: Fix integer overflows on 32-bit systems
Ensure that the integer type used in expressions involving multiplication with NSEC_PER_SEC is large enough to avoid overflows on 32-bit systems. In the expressions fixed by this patch a 64-bit type (long long) is required. Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com> Reviewed-by: Daniel Stone <daniels@collabora.com>
This commit is contained in:
parent
bd9069ffad
commit
8c919b488c
@ -160,7 +160,7 @@ ZUC_TEST(timespec_test, timespec_sub_to_nsec)
|
||||
a.tv_nsec = 1;
|
||||
b.tv_sec = 1;
|
||||
b.tv_nsec = 2;
|
||||
ZUC_ASSERT_EQ((999L * NSEC_PER_SEC) - 1, timespec_sub_to_nsec(&a, &b));
|
||||
ZUC_ASSERT_EQ((999LL * NSEC_PER_SEC) - 1, timespec_sub_to_nsec(&a, &b));
|
||||
}
|
||||
|
||||
ZUC_TEST(timespec_test, timespec_sub_to_msec)
|
||||
@ -190,7 +190,7 @@ ZUC_TEST(timespec_test, timespec_from_nsec)
|
||||
ZUC_ASSERT_EQ(1, a.tv_sec);
|
||||
ZUC_ASSERT_EQ(0, a.tv_nsec);
|
||||
|
||||
timespec_from_nsec(&a, (5L * NSEC_PER_SEC) + 1);
|
||||
timespec_from_nsec(&a, (5LL * NSEC_PER_SEC) + 1);
|
||||
ZUC_ASSERT_EQ(5, a.tv_sec);
|
||||
ZUC_ASSERT_EQ(1, a.tv_nsec);
|
||||
}
|
||||
|
Loading…
Reference in New Issue
Block a user