The concern we have is any API-version change in ARM is limited to ARM’s own resource types (subscriptions, RGs, tags etc.), and should have absolutely no impact on any other service, including the ring-0 services. At least that is the idea. Otherwise I feel there is something fundamentally wrong with the api-version work in ARM, or in how they are used in CLI tests.
We saw a couple of examples of such dependency in PowerShell tests as well, but were able to fix those. I think such as effort should be made for CLI as well.
Perhaps we can start with 1-2 impacted tests as an example and identify a pattern that can be applied to the rest of the tests. Otherwise it seems we are talking about 1-day work on ARM backend requiring 2 weeks of work on fixing CLI.
The concern we have is any API-version change in ARM is limited to ARM’s own resource types (subscriptions, RGs, tags etc.), and should have absolutely no impact on any other service, including the ring-0 services. At least that is the idea. Otherwise I feel there is something fundamentally wrong with the api-version work in ARM, or in how they are used in CLI tests.
We saw a couple of examples of such dependency in PowerShell tests as well, but were able to fix those. I think such as effort should be made for CLI as well.
Perhaps we can start with 1-2 impacted tests as an example and identify a pattern that can be applied to the rest of the tests. Otherwise it seems we are talking about 1-day work on ARM backend requiring 2 weeks of work on fixing CLI.