xref statsmodels/statsmodels#6524

cc @josef-pkt @kshedden

here's a maximum example from the unit tests

pvals = np.array([0.002, 0.004, 0.006, 0.008, 0.01 , 0.012])
ntests = len(pvals)

pvals_corrected_raw = pvals * np.arange(ntests, 0, -1)
pvals_corrected = np.maximum.accumulate(pvals_corrected_raw)
pvals_corrected

# expected array([0.012, 0.02 , 0.024, 0.024, 0.024, 0.024])
# in failing test array([0.012, 0.02 , 0.024, 0.024, 0.02 , 0.012])

the original pvals in the test is a view, but pvals_corrected_raw shouldn't be

(Pdb) pval.flags
C_CONTIGUOUS : True
F_CONTIGUOUS : True
OWNDATA : False
WRITEABLE : True
ALIGNED : True
WRITEBACKIFCOPY : False
UPDATEIFCOPY : False

@r-devulap can you ave a look?

yup, taking a look ..

Found the bug, np.maximum.accumulate results in input and output arrays being overlapped in memory in the maximum/minimum function. Added a pre-check to prevent vectorized code in such a scenario in PR #15612

I brought up this question when I submitted the patch (see comments #14867 (comment)). There are no tests covering np.maximum.accumulate. Can someone help me understand under what scenarios do we expect input and output arrays to overlap in memory to ensure we haven't missed all the use cases.