Fix Bugzilla 12623

[Hackerpilot]

No description provided.

[CyberShadow]

https://issues.dlang.org/show_bug.cgi?id=12623

Nit: please format your pulls according to http://wiki.dlang.org/Pull_Requests#Create_a_pull_request in the future.

[Hackerpilot]

I think that link should have been https://issues.dlang.org/show_bug.cgi?id=12623

[CyberShadow]

Fixed :)

[denis-sh]

Am I the only one here thinking the fact 1. a and 1.a are parsed differently looks like a compiler bug? If there is some discussion and reasoning about this behaviour it must be added to the documentation.

[Hackerpilot]

I think the intention was for UFCS to work with integer literals and for 1. to be a valid floating point literal. I personally think that this is a dumb idea, but I'm tired of the documentation being inconsistent with the compiler's behavior and this was easier than fighting another "the compiler should not accept this" fight.

[denis-sh]

We can't just document all dmd bugs and be happy. Actually we can, but shouldn't.

[MartinNowak]

Deprecating 1. as floating point literal would be hard, because it's valid C.

[ghost]

I seem to recall we have already deprecated something similar. Was it just 1.f to allow UFCS? Strange that we've left 1. work though.