Classes with the parent selector in their name – can't use them as mixins

[tomek-he-him]

Similar to #2216, but it's still there in v2.3.1 .

.a {
  &-b {
    some: stuff;
    }
  }

.c {
  .a-b;
  }

Piped through lessc, this throws a

NameError: .a-b is undefined in ... on line 8, column 3:
7 .c {
8   .a-b;
9   }

Just for the record, without the .c definition everything works as expected:

$ cat <<EOF | lessc -
> .a {
>   &-b {
>     some: stuff;
>     }
>   }
> EOF

.a-b {
  some: stuff;
}

[seven-phases-max]

This code is not supposed to work currently. Selectors with identifiers concatenated via nesting are not available as mixins. (And no, it's not the same as #2216 - selector "interpolation" there and "concatenation" here are different things).
(The open ticket for this feature-request is #1399 though it contains a lot of not so very related stuff so it not so easy to find anything there. So maybe we need a dedicated one...).

[tomek-he-him]

@seven-phases-max considering what you said in #2453 this feature wouldn't be necessary, if one could extend .a-b. But that's not possible neither at the moment (see my comment therein).

[seven-phases-max]

So feel free to close it (well, as I mentioned above such examples are given in #1399 so eventually this might be fixed/implemented as a part of more wide mixin matching improvement regardless of use-cases).