Don't require 32-bit to be time32

Debian armhf/armel (but not i386) are now time64

Author: Graham Inggs, Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/1068104
Forwarded: no

Gbp-Pq: Name 1068104_time64.patch

Don't fail sum test on near-cancelling inputs

Failed in 1.5.3+dfsg-11 after it happened to select an input that
cancelled to ~1e-5 (on i386, so x87 excess precision might be involved)

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name sum_loosen_test_tolerance.patch

Ignore DeprecationWarnings from Python 3.12

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name ignore_python3p12_deprecations.patch

Ignore pytables test failures with Python 3.12

The combined xfails are because two separate xfails where only one
has run=False may run, which is a problem when the run=False exists
to avoid a crash - see 1.5.3+dfsg-8 armhf build log

Bug-Debian: https://bugs.debian.org/1055801
Bug-Ubuntu: https://launchpad.net/ubuntu/+bug/2043895
Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name pytables_python3p12.patch

Fix test failures when xlsxwriter is not installed

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name 2p1_openpyxl_errors.patch

Mark tests that need openpyxl

(These do the usual loop through engine/read_ext but
actually use .xlsx every time)

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name allow_no_openpyxl.patch

Work around xarray wrongly reporting its version

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name xarray_version_workaround.patch

Ignore exceptions in documentation examples

Some examples download data, and/or depend on
packages Debian doesn't have, so can't run in a Debian build

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: not-needed

Gbp-Pq: Name ignore_ipython_exceptions.patch

Don't crash on clean

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name unbreak_clean.patch

Don't try to run matplotlib-using tests without it

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name allow_no_matplotlib.patch

On mips, 1**np.nan and np.nan**0 may be NaN not 1

Done this way not a plain xfail to allow only this difference,
not clearly wrong answers

(The same hardware's "invalid value encountered" warnings,
probably from sNaN/qNaN being reversed, are ignored elsewhere:
d/rules and xfail_tests_nonintel_io.patch)

https://en.wikipedia.org/wiki/NaN#Encoding

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name mips_pow_nan.patch

Warn that numba may not work on non-x86

Currently known issues are crashes, not wrong answers, but because
the test setup doesn't allow ignoring crashes but failing on
wrong answers, it would be easy to not notice if this changed

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name numba_warn_nonx86.patch

Don't require python3-tzdata

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/1043968
Forwarded: no

Gbp-Pq: Name accept_system_tzdata.patch

Don't assume little-endian in test references

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name tests_dont_assume_endian.patch

Don't fail plot tests on rounding error

(upstream seem to have _disabled_ the affected tests
...see also test_series)

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/1029251
Forwarded: no

Gbp-Pq: Name 1029251_ignore_rounding_error.patch

Don't require a warning armel numpy doesn't have

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no (this version requires dpkg)

Gbp-Pq: Name armel_ignore_nonwarning.patch

Fix test failures on 32-bit systems

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: partly https://bugs.debian.org/1026351
Forwarded: no

Gbp-Pq: Name tests_dont_assume_64bit.patch

Fix arch-specific upstream xfails

We test on more architectures, so upstream's xfails are not always
correct everywhere.  On those known to fail:
arm64 xfail -> all non-x86 xfail
x86 or unconditional strict xfail -> unconditional nonstrict xfail

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug: https://github.com/pandas-dev/pandas/issues/38921, https://github.com/pandas-dev/pandas/issues/38798, https://github.com/pandas-dev/pandas/issues/41740, https://github.com/numpy/numpy/issues/19146
Forwarded: no

Gbp-Pq: Name fix_overly_arch_specific_xfails.patch

Skip test that fails from outside the source tree

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: not-needed

Gbp-Pq: Name skip_test_missing_required_dependency.patch

Avoid test failures on Hurd

Allow multiprocessing to be unavailable
Accept any errno not just 2 for (intentionally) nonexistent files
(Hurd appears to use 2**30+2)

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name hurd_compat.patch

Allow some numba errors on 32-bit

Specifying the exception type allows only explicit errors,
not silently wrong answers

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name numba_fail_32bit.patch

Remove code from Stack Overflow

Stack Overflow content is CC-BY-SA licensed,
which this package is not supposed to be.  These snippets may be
too small to be copyrightable, but removing them to be safe.

https://lists.debian.org/debian-legal/2020/04/threads.html#00018

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no - deletes some tests/examples without replacement

Gbp-Pq: Name remove_ccbysa_snippets.patch

Allow tests to use the data files in the source tree

We don't ship these in the package,
but do want to run the tests that use them

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: https://github.com/pandas-dev/pandas/issues/54907

Gbp-Pq: Name find_test_data.patch

Link to rather than embed Google calendar

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: not-needed

Gbp-Pq: Name privacy.patch

Use fixed seeds for reproducible pseudorandomness

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: no

Gbp-Pq: Name fix_random_seeds.patch

Don't try to read a contributor list from the git log

Debian packages are built from tarballs, so there isn't a git log.

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: not-needed

Gbp-Pq: Name contributor_list_not_in_tarball.patch

Use packaged intersphinx indexes

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/876417
Forwarded: not-needed

Gbp-Pq: Name use_system_intersphinx.patch

Use Debian packaged mathjax

Author: Andreas Tille <tille@debian.org>
Forwarded: not-needed

Gbp-Pq: Name mathjax-path.patch

Avoid privacy breach by analytics

Author: Yaroslav Halchenko <debian@onerussian.com>,
        Andreas Tille <tille@debian.org>,
        Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: not-needed

Gbp-Pq: Name deb_disable_analytics.patch

HDF5 and Stata I/O are broken on some architectures

Fix some issues, warn on use and xfail tests for the remainder

Everything that has a run=False xfail in here should also be in
the run-and-ignore set in debian/tests/numbatests

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 was previously seen to fail 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
Bug: partly https://github.com/pandas-dev/pandas/issues/54396
Forwarded: no

Gbp-Pq: Name xfail_tests_nonintel_io.patch

Don't try to import from the source directory

Needed as we build the extension modules elsewhere

Author: Yaroslav Halchenko
Forwarded: not-needed

Gbp-Pq: Name deb_doc_donotoverride_PYTHONPATH.patch

Don't try to use git to find the version number

Needed as Debian buildds use tarballs

Author: Yaroslav Halchenko
Forwarded: not-needed

Gbp-Pq: Name deb_nonversioneer_version.patch

pandas (2.2.3+dfsg-7) unstable; urgency=medium

  * Re-add some but not all build-deps, stop ignoring most tests
    and documentation fails (but still ignore #1094417).
  * Temporarily move matplotlib to Build-Depends-Indep.
  * Tests: remove no longer needed xfails.

[dgit import unpatched pandas 2.2.3+dfsg-7]

Import pandas_2.2.3+dfsg-7.debian.tar.xz

[dgit import tarball pandas 2.2.3+dfsg-7 pandas_2.2.3+dfsg-7.debian.tar.xz]

Import pandas_2.2.3+dfsg.orig.tar.xz

[dgit import orig pandas_2.2.3+dfsg.orig.tar.xz]