On 5/7/26 09:05, Chen Wandun wrote: > Add madv_collapse_range to verify the fix for the spurious -EINVAL > returned by madvise_collapse() when the madvised range contains no > complete PMD-aligned window.
Do we really need 141 LOC of test to verify a "userspace is being stupid" scenario? Nothing was really broken, it's just a suboptimal return value, or what am I missing? -- Cheers, David
