functionConfig validation for eval and render
#2133
Conversation
natasha41575
changed the title
metadata.name field in functionConfigs
natasha41575
force-pushed
the
functionConfigs
branch
from
082fbc7
to
6eb82e6
Compare
natasha41575
force-pushed
the
functionConfigs
branch
from
332747a
to
7fba3c8
Compare
|
Please update the PR to cover #1959 |
natasha41575
force-pushed
the
functionConfigs
branch
from
7fba3c8
to
04ca55f
Compare
natasha41575
changed the title
metadata.name field in functionConfigsmetadata.name field in functionConfigs and share functionConfig check between eval and render
natasha41575
changed the title
metadata.name field in functionConfigs and share functionConfig check between eval and rendereval and render
natasha41575
force-pushed
the
functionConfigs
branch
from
c19edd0
to
7454436
Compare
natasha41575
force-pushed
the
functionConfigs
branch
from
7454436
to
cd0703c
Compare
|
@Shell32-Natsu please take a look when you can, thanks! |
| # limitations under the License. | ||
|
|
||
| testType: eval | ||
| image: gcr.io/kpt-fn/set-namespace:v0.1 |
There was a problem hiding this comment.
Please don't use non-concrete tag like unstable or v0.1 if it's possible. Use full version v0.1.x in kpt e2e test. So we will not suffer from break changes in function.
There was a problem hiding this comment.
@mengqiy Why are we overwriting 0.1 tags? That may have been OK earlier, but going forward 0.1 should not be treated special (otherwise what's the difference between unstable and v0.1?), and the versioning scheme should apply to it. Using 0.1 should be fine in order to pick up patch releases if needed.
There was a problem hiding this comment.
We had an issue about breaking change in v0.1 in past a few weeks. 0.1 should be stable but for kpt e2e test I think we should pin to a specific version of function.
There was a problem hiding this comment.
We should dogfood the versioning scheme we're promising our users. If we're telling them v0.1 is floating to pick up only patches, then we want to feel the pain if it breaks. Otherwise, our e2e is stable, but the user's workflow are broken. Which is worse?
fixes #1959, #2105