Repository navigation
Conversation
|
Any chance one of the maintainers could look at this? It's horribly annoying. |
|
@javiereguiluz sorry to bother you, but is there a chance for this to get merged? I've been using this patched in for... well months really without problems and it solves few issues |
GromNaN
left a comment
There was a problem hiding this comment.
Thanks for tackling this! The logic is sound: keeping the app-namespaced candidate as a fallback and preferring it when it exists is exactly what avoids the silent vendor-file write reported in #1566. I reproduced the original issue and confirmed the fix behaves as expected.
a058b83 to
5b6eaf2
Compare
I'll open an alternative PR that fixes also core classes conflicts.
|
Yeah, feel free to open up an alternate as this PR is extremely stale. I'd appreciate some sort of mention though. Once it's ready, I'll close this. |
|
This issue should be resolved by #1838 |
Fixes #1566
I've pasted this over the currently installed version in the project in which I had these issues, seems to work fine. Would be good if it could be re-checked and generally looked over, since I'm not 100% sure about it, but this is the place which causes the issue mentioned.
It should prevent situations where a known, working class name is skipped over and a vendor class is used instead, as there is no instance in which modifying vendor classes is required.
Actually this also seems to not work when it comes to classes with relations, so the target entity class (one we're adding relation to) may still be pulled from the vendor.