This PR adjusts dependencies to make PDP use modelmeta >= 0.3.0. This means it can run against the pcic_meta database migrated to revision 12f290b63791. These changes include:
Upgrading modelmeta and pdp_util versions in requirements.txt
Removing version specifications for modelmeta and pdp_util from setup.py. This was causing build problems and is unnecessary.
In addition to the changes needed for the modelmeta upgrade, there are also the following:
Fix broken Python CI build. Some changes in the underlying support for our existing build broke the tests. The self-hosted base image has been adjusted to ubuntu-18.04, and the build steps changed also to compensate the loss of support. Notes:
The base image for the CI is the same as the base image for our production and test Docker images, making them all very similar and consistent.
Function get_commitish() started raising a git exception in the Python CI build. Online advice on fixing this did not work. Instead, we have a fallback to the value "unknown" if this error occurs. It seems only to occur in the CI build.
Fix broken production and test Dockerfiles, same reason.
Fix broken Metadata hyperlink in portals. This fix should have been in a separate PR, but ... The fix itself is 4 characters; the tests to catch this type of error are more bulky.
Apply Black to setup.py, which was getting messy.
And, oops, a PR for Snyk accidentally merged into this branch. Oh well.
This PR adjusts dependencies to make PDP use
modelmeta
>= 0.3.0. This means it can run against thepcic_meta
database migrated to revision12f290b63791
. These changes include:modelmeta
andpdp_util
versions inrequirements.txt
modelmeta
andpdp_util
fromsetup.py
. This was causing build problems and is unnecessary.Demonstration
In addition to the changes needed for the modelmeta upgrade, there are also the following:
ubuntu-18.04
, and the build steps changed also to compensate the loss of support. Notes:get_commitish()
started raising a git exception in the Python CI build. Online advice on fixing this did not work. Instead, we have a fallback to the value "unknown" if this error occurs. It seems only to occur in the CI build.setup.py
, which was getting messy.