[Issue 12951] New: Not all failures to compile opDispatch should be regarded as "field not found"

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=12951

          Issue ID: 12951
           Summary: Not all failures to compile opDispatch should be
                    regarded as "field not found"
           Product: D
           Version: D2
          Hardware: x86
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P1
         Component: DMD
          Assignee: nobody at puremagic.com
          Reporter: hsteoh at quickfur.ath.cx

CODE:
------
struct S {
   int x, y, z;
}
struct T {
   S impl;
   auto opDispatch(string field)() {
      return __traits(getMember, impl, feild); // N.B. this is a typo
   }
}
void main() {
   T t;
   auto x = t.x;
}
------

Compiler output:
------
test.d(13): Error: no property 'x' for type 'T'
------

Changing "feild" to "field" fixes the problem. This is not good, because the
diagnostic is completely unhelpful -- there is no indication that the compile
error is caused by a typo in the implementation of opDispatch.

I understand that sometimes it's useful for failure to compile opDispatch to be
taken as declining field dispatch (if field=="w" then the above error message
would be appropriate), but this makes it nigh impossible to debug opDispatch,
since the real error message has been gagged.

Is there a way to gag only "no property..." errors, but leave the other errors
in opDispatch ungagged?

--