dgit.raspbian.org Git - statsmodels.git/log

Merge statsmodels (0.14.6+dfsg-3) import into refs/heads/workingbranch

Ignore errors (but not wrong results) of solve_power

scipy.special.nctdtr returns NaN when rounding would otherwise make
the result out of bounds: https://github.com/scipy/scipy/issues/25470

This has only been seen to fail the statsmodels test on armhf,
but the underlying issue also exists on amd64

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

Gbp-Pq: Name test_power_ignore_exception.patch

Remove duplicate tests

pytest now errors out on duplicate test IDs by default

Author: Kevin "bashtage" Sheppard
Origin: part of upstream f4975702
Forwarded: not-needed

Gbp-Pq: Name remove_duplicate_tests.patch

Ignore a non-converging test

(it currently does warn about possible non-convergence and
it would be better to check that it does so)

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/1079789
Forwarded: no

Gbp-Pq: Name 1079789_ignore_arm64_nonconvergence.patch

Ignore test exception on big-endian systems

failed in 0.14.1+dfsg-1

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

Gbp-Pq: Name xfail_dynamic_factor_news_bigendian.patch

Fix random seeds for reproducibility

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

Gbp-Pq: Name fix_random_seeds.patch

Allow building with Cython 3.x

(= Cython _version_ 3.x, not to be confused with the cython3 binary)

Author: Chad Fulton, Rebecca N. Palmer <rebecca_palmer@zoho.com>
Origin: mostly https://github.com/statsmodels/statsmodels/pull/8961
Bug-Ubuntu: https://launchpad.net/bugs/2043284
Forwarded: no

Gbp-Pq: Name cython3p0_compat.patch

Ignore extra warnings that may occur on mips*

Some mips* hardware seems to emit
RuntimeWarning: invalid value encountered in multiply
on multiplication by explicit np.nan, which breaks these tests.

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

Gbp-Pq: Name mips_ignore_nan_warnings.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 ignore_armel_nonwarning.patch

Allow a memory-heavy test to run out of memory

Failed in 0.14.0 on armel,armhf,mipsel (but not i386)

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

Gbp-Pq: Name xfail_out_of_memory.patch

Ignore a known failure to converge

(Seen to fail in 0.13.5+dfsg-5, twice on mipsel, in 0.14.0 on armel,
and in 0.14.6 on amd64; also known upstream.
Not a plain mark.xfail to not accept "converged" to the wrong answer.)

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: not-needed
(upstream 15f574d also ignores it)

Gbp-Pq: Name test_corr_nearest_factor_sparse_ignore_nonconverge.patch

Avoid cast exception in np.bincount

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/1030208
Forwarded: no

Gbp-Pq: Name 1030208_scipy1p10.patch

Ignore missing HessianInversionWarning on some mips*

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

Gbp-Pq: Name mips_ignore_nonwarning.patch

Loosen tolerance on a possibly unstable test

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/1020060 and 1014278
Forwarded: https://github.com/statsmodels/statsmodels/issues/8341

Gbp-Pq: Name 1020060_loosen_test_tolerance.patch

Avoid test failure on mipsel

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug: https://github.com/statsmodels/statsmodels/8341
Bug-Debian: https://bugs.debian.org/1014884
Forwarded: no

Gbp-Pq: Name 1014884_loosen_test_tolerances.patch

test_seasonal_order fails on i386

Warn on use and xfail

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

Gbp-Pq: Name xfail_i386_ets.patch

Xfail / increase tolerance on tests

TestMICE and test_mixedlm vary enough with the random state, and
corrpsd is close enough to 0, that this is plausibly a rounding non-bug

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: https://bugs.debian.org/997081
Forwarded: https://github.com/statsmodels/statsmodels/issues/7911

Gbp-Pq: Name 997081_xfail.patch

Remove code from Stack Overflow and wasteful loop

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

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

Gbp-Pq: Name remove_ccbysa_snippet.patch

Allow tests to fail if multiprocessing is not available

This is currently the case on hurd-i386

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

Gbp-Pq: Name xfail_no_multiprocessing.patch

Use local requirejs/mathjax

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

Gbp-Pq: Name sphinx_local_requirejs.patch

Be reproducible, and don't fail on examples exceptions

Don't execute example notebooks because some of them
have unreproducible output, either numerical differences (small
enough that I guess them to be machine-dependent rounding error
not deliberate randomness) or extra cells (a single newline
being split into its own cell vs part of the previous one).
Copy them instead.

At one point these 10 were affected, but skipping just them doesn't
work because it seems 'random' which ones have the extra cells.
count_hurdle
discrete_choice_example
discrete_choice_overview
exponential_smoothing (and its images 14_0 and 16_0)
generic_mle
glm_formula
glm_weights
postestimation_poisson
stats_poisson
treatment_effect

Also remove timestamps, ipykernel IDs, etc from examples output
for reproducibility. (This is here not in debian/rules to do it
before building the search index.)

Skip some vector_ar cells that would print <type at address>
output (that would end up in the search index, so
stripping it afterwards isn't enough).

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

Gbp-Pq: Name sphinx_reproducibility_and_ignore_errors.patch

Don't try to use a theme we don't have

Upstream switched to sphinx_material in
1d7aafad5c3ad634c9ae2122881c1809c2ccbcbe
and later to sphinx_immaterial; Debian doesn't have those themes.

(I haven't investigated packaging it, which might be a better long-term option.)

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

Gbp-Pq: Name no_sphinx_material.patch

Don't try to write to the source directory

Not allowed in autopkgtest

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

Gbp-Pq: Name use_tmp_path.patch

Use Python 3 in script shebangs

(Skipping the two .pyx files with #!python - they need to be compiled
with Cython, not directly run with any version of plain Python)

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

Gbp-Pq: Name python3_shebangs.patch

Loosen tolerances on some tests on i386

and xfail a test that fails for not producing ConvergenceWarning.

Failure logs:
test_smoothing
https://buildd.debian.org/status/fetch.php?pkg=statsmodels&arch=i386&ver=0.9.0-3&stamp=1567157609&raw=0
test_multivariate_switch_univariate
https://buildd.debian.org/status/fetch.php?pkg=statsmodels&arch=i386&ver=0.12.0-1&stamp=1599693472&raw=0
TestMixedLM.test_compare_numdiff
salsa-ci for 0.14.2-1

As it only fails on i386 and isn't far wrong,
I suspect different rounding due to x87 excess precision

Author: Rebecca N. Palmer <rebecca_palmer@zoho.com>
Bug-Debian: partly https://bugs.debian.org/938949
Forwarded: no

Gbp-Pq: Name i386_loosen_test_tolerances.patch

Use example data from an R package we have

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

Gbp-Pq: Name use_available_data.patch

Use intersphinx inventory files from other Debian packages.

Forwarded: not-needed

Gbp-Pq: Name use_system_inventory.patch

Use cached datasets in tests and examples

Also remove a download that isn't actually used in that example.

This allows the tests to pass, and at least some of the examples to
be built, in an offline environment such as a Debian buildd.

The cached data is extracted from R packages by debian/datasets/*.

Author: Diane Trout <diane@ghic.org>, Rebecca N. Palmer <rebecca_palmer@zoho.com>
Forwarded: not-needed

Gbp-Pq: Name use_cached_datasets.patch

statsmodels (0.14.6+dfsg-3) unstable; urgency=medium

  * Tests: ignore explicit failure of solve_power,
    probably caused by rounding issues.
  * Salsa CI: test clean, don't try to build on i386.

[dgit import unpatched statsmodels 0.14.6+dfsg-3]

Import statsmodels_0.14.6+dfsg-3.debian.tar.xz

[dgit import tarball statsmodels 0.14.6+dfsg-3 statsmodels_0.14.6+dfsg-3.debian.tar.xz]

Merge statsmodels (0.14.6+dfsg-1) import into refs/heads/workingbranch

Import statsmodels_0.14.6+dfsg.orig.tar.xz

[dgit import orig statsmodels_0.14.6+dfsg.orig.tar.xz]