Purge consolidated protocols before compilation #406
Conversation
| Enum.map(beams, &(&1 |> Path.rootname(".beam") |> String.to_atom() |> purge_module())) | ||
| end | ||
|
|
||
| # Code.delete_path(path) |
There was a problem hiding this comment.
Is there a reason not to delete the path here? It seems like it is because we don't use Code.prepend_path, so I guess the question is what purpose is Code.prepend_path serving in the phoenix code reloader. I don't feel like I personally understand why Code.prepend_path is needed in the Phoenix case, so I'm not sure if we would need it or not.
There was a problem hiding this comment.
I don't know that either but here Code.delete_path(path) doesn't seem to make any difference. BTW after compile it is prepended again.
There was a problem hiding this comment.
Okay, in that case maybe we just leave a comment explaining the reasoning.
|
For future reference, here is the ecto issue that initiated the discussion: elixir-ecto/ecto#3464 |
|
any chance this could be released :)? |
|
Yeah, good idea. I'll plan to cut a release this weekend. |
|
@dvic FYI, just released 0.6.3 which includes this fix 👍 |
I was able to isolate this to a test case. It turns out that this is only evident with dialyzer on. The issue is triggered by module load in
elixir-ls/apps/language_server/lib/language_server/dialyzer/success_typings.ex
Line 22 in 3acc1fd
As pointed out earlier this needs a refactor (#407)
I applied a solution similar to phoenixframework/phoenix@b5580e9
Fixes #395