Function calls are parsed as constructors about glsl4idea HOT 6 CLOSED

@Darkyenus this is fairly serious :(

from glsl4idea.

Okay, it looks like the cause is that GLSLParsing#parseFunctionIdentifier tries to parse a type specifier, and if it can then it assumes it's a constructor. But a single identifier is a valid type specifier (it's a reference to an existing struct type).

I think given this there's probably no sensible way to separate constructors and functions at the parsing level - that distinction should come higher up, when a function/constructor call would resolve to either a function declaration or struct name (and the return type would be based on that).

from glsl4idea.

Noticed during prerelease testing as well, fixed in 5680358. It is not ideal, since structures can have constructors as well.

from glsl4idea.

Yeah, I'll look into that later, for now it should work well enough. The problem is, that it needs a reliable way to tell whether the name of function call corresponds to defined structure. Is that possible? This seems vaguely related to your work around rename refactoring.

from glsl4idea.

I'm going to be doing even more work on references - it should be possible to reliably resolve the name to either a struct type or a function name.

from glsl4idea.

When I tested it now on master, the struct name refactoring didn't work yet, but valid code involving struct constructor was not highlighted as error. So I won't postpone the release because of it and the constructor related code can be refactored back into function calls in next release.

from glsl4idea.