Validate Qualifiers at compile time #645
Conversation
|
We should try to keep a PR focused on it's change only and not include other unrelated changes into it. My instinct is that "fix event publishers not getting detected" and "refactor reading external modules" should have been in separate PRs? One aspect of this is that the PR Title is important documentation, and the other is from a git history perspective we want a commit to be focused on it's change. Putting multiple things into a commit makes life harder for our future selves when we try to understand the interaction between features by looking at commit history. |
This reverts commit 96c3a94.
sure I can break the external one into it's own thing, but the event publisher bug only arises from the changes in this PR. I'm totally unable to replicate it when I try it on master. |
Ok cool. Then lets keep that in as part of this PR. |
Fixes a regression that breaks partial compilation when qualifiers are used. Regression introduced in 10.1 via: https://github.com/avaje/avaje-inject/pull/645/files With this regression we observe compile errors stating that dependencies with qualifiers are missing. When reviewing the relevant generated @metadata dependsOn attributes in the generate source we see that there are qualifiers incorrectly appended to the factory type. The fix is in the MetaData constructor, such that the dependsOn does NOT use the name/qualifier of the type.
* Fix partial compile regression introduced in 10.1 via #645 Fixes a regression that breaks partial compilation when qualifiers are used. Regression introduced in 10.1 via: https://github.com/avaje/avaje-inject/pull/645/files With this regression we observe compile errors stating that dependencies with qualifiers are missing. When reviewing the relevant generated @metadata dependsOn attributes in the generate source we see that there are qualifiers incorrectly appended to the factory type. The fix is in the MetaData constructor, such that the dependsOn does NOT use the name/qualifier of the type.
It seems previously we only validated that all the bean types were in the scope at compile time without looking at the qualifiers.
resolves #644