The sbuild task reports "success" for packages that cannot be built on the requested architecture (Architecture: i386 vs. building on arm64)
Describe the bug
Observed in work request 43135 (sbuild task).
I scheduled a build of i810switch on arm64. The problem being that i810switch cannot be build on arm64 given its Architecture: amd64 hurd-i386 i386 kfreebsd-amd64 kfreebsd-i386
(arm64 vs. amd64).
The result was a "success" according to the debusine API and diving into the logs, we see:
[...]
E: dsc: arm64 not in arch list or does not match any arch wildcards: i386 kfreebsd-i386 amd64 hurd-i386 kfreebsd-amd64 -- skipping
aborted: False
returncode: 0
[...]
This is a mismatch in expectations, since this test rebuild have not validated that I can build i810switch in my changed environment. In a bulk test setting, I would likely have overlooked this as a problem since I would have trusted "success" to mean "Successfully rebuilt in the modified environment", not "Successfully did nothing, and you should schedule the build correctly next time".
Side note: Colin suggested that this problem would not affect sbuild
workflows, only sbuild
tasks.
How to reproduce the bug
Schedule a sbuild
task with a package that has an architecture restriction (such as i810switch) to have it be build for an architecture that is not in the architecture requested (DEB_HOST_ARCH
).
Runtime environment
It happened on debusine.d.n. The rebuild was in sid with a rebuild dpkg
. The modified dpkg
should not be affecting the outcome, since it is the debusine
+ sbuild
side that causes this (already before the build can start).
Operating system
It happened on debusine.d.n.
Versions of debusine and its dependencies
I was using 0.6.0
of the debusine
client.