Open FissoreD opened 5 days ago
However, I'm not completely sure that the second implementation will work in all cases.
I think there could be two non-unfiable terms (for elpi at least) that however unify for coq.
For example, there could exists two terms $t_1$ and $t_2$ assigned to $A6$ and $A4$ such that $t_1 \neq t_2$ in elpi
where coq.unify-eq A4 A6
would actually succeed
The compilation of
ucmra_unit_left_id
in iris gives the following elpi ruleNote that the type of
ucmra_unit_left_id
isall the
ucmra_car A
could be replaced with a uniquecoq.unify-eq
The example could produce a elpi rule like: