Repository navigation
Resolve how to handle coercions and the like under default binding modes #44848
Description
One of the tricky questions that came up when implementing default binding modes is how to handle the compiler's current logic around patterns with ref bindings. Currently, if a ref-binding is included, this disables coercion for let statements and tweaks how we handle match statements. It's unclear precisely what we ought to do now that ref bindings are not always syntactically visible. This issue exists to track the question.
There has already been a number of conversations on the topic. Here are some links for background reading:
- @nikomatsakis commenting on why we can probably get away with only considering syntactic bindings.
- @arielb1's response points out that we might want to consider possible future behavior around
DerefPureas well - Some follow-up discussion between @nikomatsakis and @arielb1 on gitter
- A very early conversation on the topic.
Activity
Auto merge of #49394 - cramertj:stable-match-bindings, r=nikomatsakis
Auto merge of #49394 - cramertj:stable-match-bindings, r=nikomatsakis
nikomatsakis commentedon Jan 17, 2019
The current handling seems to be adequate and -- in any case -- is stabilized. We permit coercions unless there is a
refbinding. This is ok because coercions don't alter things underneath references, generally speaking. Closing.