Add failing tests for exposing the value of .Self through an interface #5957
Conversation
|
I got a bad feeling I am getting the typish levels wrong here after continuing to stare at things. |
danakj
changed the title
danakj
force-pushed
the
test-call-returning-facet-value
branch
from
f75b1f7 to
96a1ffd
Compare
|
Thanks for that review, I removed the bogus test, and I have added a bunch more tests of .Self that I was using to test #5966 while working on it. It would be good to make sure we agree they are all supposed to pass as written. |
danakj
force-pushed
the
test-call-returning-facet-value
branch
from
38728ea to
eecce58
Compare
danakj
force-pushed
the
test-call-returning-facet-value
branch
from
eecce58 to
fd9072a
Compare
|
Updated the nesting tests with what we discovered in open discussion today. |
|
bump for review |
danakj
force-pushed
the
test-call-returning-facet-value
branch
2 times, most recently
from
5758d72 to
cb85ed0
Compare
danakj
force-pushed
the
test-call-returning-facet-value
branch
from
cb85ed0 to
afc3467
Compare
| // Shows both `I(.Self)` are `I(T)`. | ||
| // CHECK:STDERR: fail_todo_nested_period_self.carbon:[[@LINE+4]]:9: error: found cycle in facet type constraint for `.(I(T).A)` [FacetTypeConstraintCycle] | ||
| // CHECK:STDERR: T as (I(T) where .A = (I(T) where .A = {} and .B = {})); | ||
| // CHECK:STDERR: ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
| // CHECK:STDERR: | ||
| T as (I(T) where .A = (I(T) where .A = {} and .B = {})); |
There was a problem hiding this comment.
I don't think this should actually be a cycle, but it's still being reported as one when I get .Self working more, so pointing this out to double check.
There was a problem hiding this comment.
Yeah, I agree this is a false positive. I guess our cycle checking is incorrectly thinking the .A nested inside the second where is a rewrite for the outer .A, or something like that?
| // CHECK:STDERR: U as (I where .I1 = U); | ||
| // CHECK:STDERR: ^~~~~~~~~~~~~~~~~~~~~~ | ||
| // CHECK:STDERR: | ||
| U as (I where .I1 = U); |
There was a problem hiding this comment.
Do we have tests of the opposite condition somewhere too?
U as (I where .I1 = .Self);
(U as type) as (I where .I1 = .Self);
| // CHECK:STDERR: u.(I.G)(); | ||
| // CHECK:STDERR: ^~~~~~~ | ||
| // CHECK:STDERR: | ||
| u.(I.G)(); |
There was a problem hiding this comment.
Would it make sense to split this and the next line out from the file that contains the previous let declaration, so that we don't have a mixture of "should not fail but does" and "should fail and does" in the same file?
| // Shows both `I(.Self)` are `I(T)`. | ||
| // CHECK:STDERR: fail_todo_nested_period_self.carbon:[[@LINE+4]]:9: error: found cycle in facet type constraint for `.(I(T).A)` [FacetTypeConstraintCycle] | ||
| // CHECK:STDERR: T as (I(T) where .A = (I(T) where .A = {} and .B = {})); | ||
| // CHECK:STDERR: ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
| // CHECK:STDERR: | ||
| T as (I(T) where .A = (I(T) where .A = {} and .B = {})); |
There was a problem hiding this comment.
Yeah, I agree this is a false positive. I guess our cycle checking is incorrectly thinking the .A nested inside the second where is a rewrite for the outer .A, or something like that?
| // TODO: Improved error: no member named `I1` in `T` of type `type`. | ||
| // |
There was a problem hiding this comment.
I think this ought to be valid: because T is a template parameter, we ought to defer the lookup until we have a non-template value for it.
We test using .Self in a generic parameter of an interface and as the value of an associated constant.