Merge lp:~aaron-whitehouse/duplicity/launchpad_tox_profile into lp:~duplicity-team/duplicity/0.7-series
Proposed by
Aaron Whitehouse
| Status: | Merged |
|---|---|
| Merged at revision: | 1121 |
| Proposed branch: | lp:~aaron-whitehouse/duplicity/launchpad_tox_profile |
| Merge into: | lp:~duplicity-team/duplicity/0.7-series |
| Diff against target: |
52 lines (+16/-6) 3 files modified
README-REPO (+7/-4) po/duplicity.pot (+1/-1) tox.ini (+8/-1) |
| To merge this branch: | bzr merge lp:~aaron-whitehouse/duplicity/launchpad_tox_profile |
| Related bugs: |
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| duplicity-team | Pending | ||
|
Review via email:
|
|||
Commit message
Add tox testing profile that mimics the packages installed on the Launchpad build server, to reduce the likelihood of tests passing our test suite, but failing on the build server (e.g. because of the out-of-date mock version).
Description of the change
Add tox testing profile that mimics the packages installed on the Launchpad build server, to reduce the likelihood of tests passing our test suite, but failing on the build server (e.g. because of the out-of-date mock version).
To post a comment you must log in.
Revision history for this message
| Aaron Whitehouse (aaron-whitehouse) wrote | # |
Preview Diff
[H/L] Next/Prev Comment, [J/K] Next/Prev File, [N/P] Next/Prev Hunk
Note that this change does introduce another test run. On my machine, each run takes about 5 minutes, so run-tests has increased from taking 10 minutes to 15 minutes. It does, however, make it very clear if any test failures result from Python 2.6, 2.7 or the particular set-up on the build server.
To me, this is probably the best approach. 15 minutes does not seem ridiculous for a final pre-commit test before proposing a merge. If a developer is doing a lot of testing to isolate and fix a bug, they can use the tox commands set out in the README-REPO to test a narrower range of tests, or to only test against one or two test environments (py26, py27 or lpbuildd).
If you disagree, however, and think that we need to limit the testing to two runs, there are two obvious options to me:
1) Change the command in testing/run-tests to be "tox -e py26,py27", so that everything is as it was before and the new lpbuildd profile is only used if "tox" or "tox -e lpbuildd" is run (likely by one of us looking for a build issue).
2) Change the command in testing/run-tests to be "tox -e py26,lpbuildd". The Launchpad build server runs Python 2.7 anyway, so testing against Python 2.6 and the Launchpad build server profile should find any issues with either Python version. The key downside to this approach is that if there are any issues, such as the recent issue with Launchpad's mock.patch not taking a return_value argument, it would be harder to determine whether the issue was caused by the Python version or the obsolete build server package versions.