Xfail NaN <-> NaT tests on non-x86 and warn on cast
pd.Series([np.nan]).astype('datetime64[ns]')[0] = pd.NaT on x86
but 1970-01-01 on arm* because float NaN -> int is undefined:
https://github.com/numpy/numpy/issues/8325
https://github.com/pandas-dev/pandas/issues/17792
https://github.com/pandas-dev/pandas/issues/26964
On s390x it's the maximum _positive_ value (2**63-1 ns = year 2262)
On riscv64 one test case raises an exception
(though I suspect not the general case since there aren't more).
Author: Andreas Tille <tille@debian.org>, Graham Inggs <ginggs@debian.org>, Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/877754
Forwarded: no
HDF5 and Stata I/O are broken on some architectures
Fix some issues, warn on use and xfail tests for the remainder
armhf TestHDF5Store::test*encoding only sometimes crashes
(1.1.3+dfsg-1 passed on build but failed autopkgtest)
HDF5 and Stata are known to fail on big-endian architectures
Stata also fails on qemu-ppc64el, but not real ppc64el
Author: Andreas Tille <tille@debian.org>, Graham Inggs <ginggs@debian.org>, Yaroslav Halchenko <debian@onerussian.com>, Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/877419
Forwarded: no