xsimd/complex/batch errors are resolved by xtensor 0.24 with
xsimd 8.1.0-3.

Closes: #994246, #994248, #1013625.

Docs are actually generated by sphinx not doxygen, so mathjax patch
needs to be applied to conf.py.

Keep the Doxygen patch in any case.

Build-Depends-Indep: libjs-mathjax

This reverts commit f97c7927.

Reinstate docs_mathjax.patch.

This reverts commit dacf0b22.

patch doesn't work, lintian still whines

W: xtensor-doc: privacy-breach-generic [<script async="async" src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js">] (https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js) [usr/share/doc/xtensor-doc/html/random.html]

configures docs to use local mathjax files

applied upstream

Update to upstream version '0.24.2'
with Debian dir 15956fcad565490c1a6c8780b7cbf0f03736aa03

applies upstream PR#2462 to fix installation of headers

doctest_PR2405.patch
test_EXPECT_DOUBLE_EQ_PR2436.patch
doc_sphinx_API.patch

Update to upstream version '0.24.0'
with Debian dir acf46995575565530f3e602fd41e288ee04e14ee

from the uscan manpage

run clang tests from test-xtensor-clang to get name in CI test log

enables failure of xreducer.average test to be ignored if
MARK_FLAKEY_TEST is defined.

Used to enable i386 to build.

Mark test as "may_fail" rather than "skip" to enable the underlying
bug to be monitored.

manage low-memory arches using "-Os -g0"

manage mipsel by switching off -march=native
(cmake -Darch_native_supported=OFF)
-mxgot is not needed in that case.

Thanks YunQiang Su and Adrian Bunk for help resolving the mipsel build
problems.

remove -g build flag on low-memory arches (required for mipsel to
succeed, with or witout -O2 or -O0)

Closes: #995867.

That's amd64 only (and kfreebsd is untested).

Restrict in debian/tests also. arm64, i386 tests continue to fail
(Bugs #994246, #994248)

Doesn't help mipsel, and only makes the i386 build slow.

provides cmake option BUILD_TEST_XNPY to skip test_xnpy in tests.

Required to build s390x, hppa, powerpc, ppc64, sparc64.

Default BUILD_TEST_XNPY=ON for other arches.

See Bug#985699, and
https://github.com/xtensor-stack/xtensor/issues/2328

needed to avoid virtual memory exhausted.

Use together with --no-parallel.

to ensure build independent from network

updates sphinx API in conf.py to fix doc build with sphinx 4.2.

see https://github.com/xtensor-stack/xtensor/issues/2441
https://github.com/sphinx-doc/sphinx/issues/7747