Source: blends
Version: 0.7.10
Severity: normal
Tags: ftbfs patch
User: debian-qa@lists.debian.org
Usertags: apt-cache
Hi,
blends uses the apt cache during testing. This breaks on reproduce.d.n
as it has an empty apt cache:
https://reproduce.debian.net/all/#blends-common
The attached patch avoids running the test in that case.
In addition the sbuild team plans to clean the apt cache before building
the package by default in forky which would make this package FTBFS.
This would also be fixed with the patch.
Cheers Jochen
diff --git a/debian/rules b/debian/rules
index 478475a..99c34ec 100755
--- a/debian/rules
+++ b/debian/rules
@@ -17,10 +17,12 @@ override_dh_auto_build:
override_dh_auto_test:
ifeq (,$(filter nocheck,$(DEB_BUILD_OPTIONS)))
+ifneq (,$(wildcard /var/lib/apt/lists/*Packages))
pytest-3 --doctest-modules
rm -f education # clean up test remnants
find . -type d -name __pycache__ | xargs rm -rf
endif
+endif
override_dh_installchangelogs:
for pkgnews in $(DEB_SOURCE)-common $(DEB_SOURCE)-dev ; do \
The provided patch just disables the build time tests for the package,
which seems not a good solution to me.
Also changing Trixie sbuild at
this stage seems a bit contradicting to the goal of the current Trixie
hard freeze, not?
I'd propose to keep this as-is for Trixie.
After its release, couldn't
we just add a "apt update" to re-fill the cache before running the
tests? Or does this already count as "network access"?
The problem is that the test explicitly opens the apt cache and I did
not find a way to mock that in a doctest. We could convert the tests
to unit tests and use mock then.
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 491 |
Nodes: | 16 (2 / 14) |
Uptime: | 127:22:04 |
Calls: | 9,688 |
Calls today: | 4 |
Files: | 13,728 |
Messages: | 6,177,249 |