Survive inaccessible types when computing implicit scope #21589
Conversation
Also: Give a better error message later when encountering a missing type that refers to a private member of a base class. The previous one was misleading since it referred to a potentially missing class file, which is certainly not the case here. Fixes scala#21543
| |Java defined annotations don't have an exact constructor representation | ||
| |and we previously relied on the order of the fields to create one. | ||
| |Java defined annotations don't have an exact constructor representation | ||
| |and we previously relied on the order of the fields to create one. |
There was a problem hiding this comment.
This PR will fail because of the changes here, some of the checkfiles will not match anymore, they need to be updated.
tests/neg/i21543.scala
Outdated
|
|
||
| new Scope { | ||
| val commands = List( | ||
| Cmd(List("1", "2")) |
There was a problem hiding this comment.
| Cmd(List("1", "2")) | |
| Cmd(List("1", "2")) // error // error |
|
The trailing spaces were removed automatically in my commit since my editor removes them. To avoid that problem in the future, we should have a commit hook that tests that there are no trailing spaces or that removes them. We used to have something like this, as far as I recall. |
|
Since I was just re-reading the forum topic on multiline strings as Java "text blocks", which also strip trailing incidental white space, I will vote for that feature with my time. Even with text blocks, it's useful to normalize trailing white space and also whether source code has a trailing newline, since the extraneous diff is a distraction. |
Also: Give a better error message later when encountering a missing type that refers to a private member of a base class. The previous one was misleading since it referred to a potentially missing class file, which is certainly not the case here.
Fixes #21543