staging-next 2022-08-06

[vcunat]

• workflow docs: nixos.org/manual/nixpkgs/unstable/#submitting-changes-commit-policy
• important job(s): hydra.nixos.org/job/nixpkgs/staging-next/unstable#tabs-constituents
• jobset: hydra.nixos.org/jobset/nixpkgs/staging-next
• nix-review reports: https://malob.github.io/nix-review-tools-reports/
• previous staging-next: https://github.com/NixOS/nixpkgs/pull/183748

[lovesegfault]

Highlight reel:

commit	description
c22150c09ef	python310Packages.jsonschema: 4.7.2 -> 4.9.1
b55345f1314	cairo: pull upstream fix for grayscale aliasing bug
6f8fdc5b87b	SDL: add patch for CVE-2022-34568
19c5c57e72d	libtiff: add patch for CVE-2022-34526
67dece3b2c0	python310Packages.attrs: 21.4.0 -> 22.1.0
b34a4df8574	[staging] cmake: 3.23.3 -> 3.24.0
83daa6a3161	nspr: 4.34 -> 4.34.1
c817efe660f	gcc: extend stripping of .a libraries and .o objects
b4ba441d7c8	python310: 3.10.5 -> 3.10.6
112cd59c6ab	vim: 9.0.0115 -> 9.0.0135
ae26091d373	python3Packages.django_3: 3.2.14 -> 3.2.15
3d17b4c305c	cmake/setup.sh: allow for cmakeBuildDir to be configured
09c61607460	libcap: 2.63 -> 2.65
5e2471dfecf	perldevel: 5.35.9 -> 5.37.0, perl-cross: 31dac3e2 -> c8760457
20868cb2d07	perl: 5.34.1 -> 5.36.0
eba5f13c08c	go_1_18: 1.18.4 -> 1.18.5
81163361cc3	openh264: 2.2.0 -> 2.3.0
8907016b22e	linuxHeaders: 5.18 -> 5.19
f2298fcf8f5	unbound: 1.16.0 -> 1.16.2
c78ef1c764f	libdvdread: 6.1.2 -> 6.1.3 (#184291)
d1cc42ef524	libnl: 3.5.0 -> 3.7.0
964ee6f55f2	libwebp: 1.2.2 -> 1.2.3
549e08c8e8a	bintools-wrapper, cc-wrapper: avoid invalid export of 'expand-response-params'
5dfbe0ac192	iwd: 1.28 -> 1.29
a4cfeb7cf18	cmake: 3.23.2 -> 3.23.3
c826b18a286	patchelf: 0.14.5 -> 0.15.0
d82ce045579	windows.mingw_w64: enable stripping
51d1d82cc1c	libass: 0.15.2 -> 0.16.0
4c1987ddb18	python310Packages.urllib3: 1.26.10 -> 1.26.11
d338b732615	ell: 0.51 -> 0.52
bf817382e73	gnupg: 2.3.6 -> 2.3.7
43f0842a919	freeglut: 3.2.1 -> 3.2.2
e384d91be48	libfido2: depend on pcsclite to support nfc keys
a79c8da189e	libfido2: 1.10.0 -> 1.11.0

[fabianhjr]

Built and booted into staging-next, required two small patches to disable 2 test cases from 2 different python3 packages. On #185627

[vcunat]

• [x] python3Packages.ipython tests are broken, portably: https://hydra.nixos.org/build/186605987 https://hydra.nixos.org/build/186574704 https://hydra.nixos.org/build/186577229 https://hydra.nixos.org/build/186537730 maintainers: @bjornfor, @FRidh. (not sure why thousands of packages depend on it transitively)
• [x] perl536Packages.CpanelJSONXS https://hydra.nixos.org/build/186584479 https://hydra.nixos.org/build/186513584 (builds on *-darwin) Perhaps /cc @stigtsp?

[bjornfor]

The IPython failure:

=================================== FAILURES ===================================
_______________________ test_pprint_heap_allocated_type ________________________

    def test_pprint_heap_allocated_type():
        """
        Test that pprint works for heap allocated types.
        """
        module_name = "xxlimited" if sys.version_info < (3, 10) else "xxlimited_35"
        xxlimited = pytest.importorskip(module_name)
        output = pretty.pretty(xxlimited.Null)
>       assert output == "xxlimited.Null"
E       AssertionError: assert 'xxlimited_35.Null' == 'xxlimited.Null'
E         - xxlimited.Null
E         + xxlimited_35.Null
E         ?          +++

IPython/lib/tests/test_pretty.py:146: AssertionError

Given the if sys.version_info < (3, 10) I'd say this test started failing when python-3.10 was made the default. Disable test?

[bjornfor]

Re: IPython test failure: https://github.com/ipython/ipython/pull/13714

[vcunat]

So in our case due to b4ba441d7c8 most likely.

[vcunat]

Applying that makes it all pass, so I trust that it's a good solution and I'm pushing that.

[stigtsp]

perl536Packages.CpanelJSONXS

Fixing failing perlPackages from perl 5.36.0 bump:

• https://github.com/NixOS/nixpkgs/pull/186212

[vcunat]

Here's some choice of more problematic dependencies with counts from nix-review-tools report. Only those regressing since master, verified locally on x86_64-linux (but usually regressed cross-platform):

• [x] python310Packages.loguru (166)
• [x] python310Packages.cattrs (72) https://github.com/python-attrs/cattrs/pull/295
• [x] libsForQt5.mailcommon (48)

---

In nix-review-tools reports it can be confusing that builds fixed by a restart are still shown as failed. And running a new eval after success apparently doesn't help, probably due to some caching. Also for staging-* it would be perhaps better to show diff against target branch.

[vcunat]

I forgot to notify maintainers.

• loguru: @jbaum98, @rmcgibbo
• cattrs: @fabaff
• mailcommon: @ttuegel, @vandenoever, @nyanloutre

[vcunat]

I don't think libsForQt5.mailcommon is a channel blocker, so maybe it's OK-ish to let it regress on master?

[vcunat]

That failure was triggered by cmake update in PR #185287. (though that information doesn't seem very useful)

[vcunat]

Uh, so I worked around the failure for now.

[NickCao]

I think I just found the missing piece of the puzzle: https://invent.kde.org/pim/messagelib/-/commit/6eaef36d42bdb05f3412c178e53077fb6224ff80

[vcunat]

I haven't tried that due to qt6 not being involved in our case.

[NickCao]

Opened https://github.com/NixOS/nixpkgs/pull/186967 to test it out.