Merge ~ahasenack/ubuntu/+source/sbuild:disco-fix-dep8-issues into ubuntu/+source/sbuild:ubuntu/devel
Status: | Merged | ||||||||
---|---|---|---|---|---|---|---|---|---|
Approved by: | Andreas Hasenack | ||||||||
Approved revision: | f94bf6f470726de1792abaf4c2249c28f1f9b806 | ||||||||
Merged at revision: | 63b0c8667bf8f6964bccfa680fc47d4db81d59eb | ||||||||
Proposed branch: | ~ahasenack/ubuntu/+source/sbuild:disco-fix-dep8-issues | ||||||||
Merge into: | ubuntu/+source/sbuild:ubuntu/devel | ||||||||
Diff against target: |
41 lines (+10/-5) 3 files modified
debian/changelog (+8/-0) debian/tests/build-procenv (+1/-4) debian/tests/control (+1/-1) |
||||||||
Related bugs: |
|
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Christian Ehrhardt (community) | Approve | ||
Canonical Server Core Reviewers | Pending | ||
Review via email: mp+360079@code.launchpad.net |
Description of the change
Bileto ticket, associated ppa, and dep8 tests:
https:/
This addresses the following issues in the sbuild DEP8 tests:
a) armhf tests
The armhf tests have been failing forever[1] because debootstrap is told to use a directory in /tmp, and the way /tmp is handled in lxd containers prevents device nodes from being created there. I changed d/t/control to require a vm for these tests. Another option would be to tell debootstrap to use another directory, perhaps something in /var/lib. This could also be a follow-up branch.
b) host release versus chroot release
The way the tests are run means their result is only valid in the development version of ubuntu at any given time. That's because they create a chroot for the development version of ubuntu, build procenv (a package) in there, and then try to install it on the *host*. This only makes sense if the host is also running the development version of ubuntu; all other cases are void and might fail. This test is a sort of time bomb as it is, because it's valid for devel, but will likely fail when SRUs start to kick in.
This is what is failing the i386 sbuild tests on bionic currently[2]
The change I made is to basically uncomment the original code that checked for this situation.
Better fixes can be made for both (a) and (b). The first one I outlined earlier, and a better fix for the second case would be to build the test package on both releases of ubuntu (host and -devel), and try to install the correct one, instead of always the devel one. I decided to leave that for another time and focus on fixing the existing test instead of adding a new one. But let me know your thoughts.
1. http://
2. http://
Ack to isolation-machine and it might even work that way.
But to be safe against future "strictness" [1] defines it as comma separated.
So I'd think we should do so on upload.
No need to overengineer (b), ack to re-install the old check for now.
+1 under the constraint to fix up or explain-why-not the comma in the Restrictions line.
[1]: https:/ /people. debian. org/~mpitt/ autopkgtest/ README. package- tests.html