Necessary for writing involved tests since the assumptions made by the
cache are violated and thus cause problems.
To keep the tests passing the oscrc intended to be used must be set as
environment variable so when osclib.conf.Config is use the config is not
lost. This would be made simpler without the needless extra config added
by StagingWorkflow.
If the init function fails due to exceptions, python still
shows exceptions caught in the __del__ function, which can
be very confusing. So avoid the destructor if the constructor
failed by setting the variables early on and return in the
destructor if self.api is not yet defined
I tried to replace the Mocking of functions with httpretty, but
unfortunately httpretty replaces all other API calls with 404s then.
What we'd need is a library sniping out single requests - possibly
by mocking the http_GET function, but that's tricky as well with
caching already overwriting it
Don't pass regexps but real URLs, otherwise the matching won't work.
On travis and leap 15.1 we have httpretty 0.9.5 which won't uncover
the problem, but on Tumbleweed we have 0.9.6 that shows it
If it's an empty project only we're freezing, the backend 'remembers' the
mtime of previous versions of the project (delete only marks it for undelete,
so creating a new empty project is just deleting this mark and as such doesn't
change the mtime).
So we need to have a new wine package with fresh content to change mtime
during freeze
The docker-compose file starts an OBS instance (on 15.0 atm) and runs
the test suite in a test container (tumbleweed), so all tests run in
an isolated environment accessing OBS at api:3000
The config default for pseudometa package (00Meta) does not
exist for Tumbleweed, so we need to add an override for :ARM
to contain the proper information so that the publishing
logic doesn't stop. Also update correspondingly for Leap
We can't generally assume ISOs can be fetched from backend (we can't
have this on IBS), so publish it - and disable the actual publishing
on the staging backend of OBS
Submitters complain more and more about the spam they're getting from
staging projects - and we rather leave that weapon for actual feedback.
One especially annoying case is if a package is added to one staging project
and then later moved to another, the submitter will receive notifications
of all kind of bots and actions for both staging projects.
Distinct copyrights were left as I do not wish to track down commit
history to ensure it properly documents the copyright holders. Also left
non-GPLv2 licenses and left bs_copy untouched as a mirror from OBS.
Already have a mix of with and without headers and even OBS does not place
on majority of files. If SUSE lawyers have an issue it will come up in
legal review for Factory.
Given the slow migration of everything to the shared project config in
tools that change project contexts multiple times a consistent way of
accessing the config without forcing remote calls is needed.
