Merge lp:~cmiller/desktopcouch/getport-at-call-time into lp:desktopcouch
Status: | Merged | ||||
---|---|---|---|---|---|
Approved by: | Elliot Murphy | ||||
Approved revision: | 44 | ||||
Merged at revision: | not available | ||||
Proposed branch: | lp:~cmiller/desktopcouch/getport-at-call-time | ||||
Merge into: | lp:desktopcouch | ||||
Diff against target: | None lines | ||||
To merge this branch: | bzr merge lp:~cmiller/desktopcouch/getport-at-call-time | ||||
Related bugs: |
|
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Elliot Murphy (community) | Approve | ||
Eric Casteleijn (community) | Approve | ||
Review via email: mp+10105@code.launchpad.net |
This proposal supersedes a proposal from 2009-08-13.
Commit message
Use subprocess.Popen and ourselves to the wait()ing, since subprocess.call() is buggy. There's still an EINTR bug in subprocess, though.
Occasionally stop couchdb in tests, so we exercise the automatic starting code. This will lead to spurious errors because of the aforementioned subprocess bug, but it's the right thing to do.
Abstract away some of the linuxisms and complain if we're run on an unsupported OS.
Fix a race condition in the process-testing code.
Replace the TestCase module with one that doesn't complain of dirty twisted reactors.
Add a means of stopping the desktop couchdb daemon.
Add an additional check that a found PID and process named correctly is indeed a process that this user started, so we don't try to talk to other local users' desktop couchdbs.
Tests run, code changes look good to me.