nim.cfg danger/release flags honored inconsistently

[disruptek]

I'm writing a benchmark tool and I thought I would run it against the #12152 test. That code takes 13s without flags and 2.7s in -d:release and 2.0s in -d:danger.

When I set -d:danger in my nim.cfg, the compiler reports that it built a Dangerous Release Build but the code takes 13s. Indeed, the binary output is identical to that of a debug build.

I've verified that the define is actually set and may be acted upon in the code with a when.

Nim Compiler Version 1.0.99 [Linux: amd64]
Compiled at 2019-10-02
Copyright (c) 2006-2019 by Andreas Rumpf

git hash: 244657035050c2b513587de45bebfefa553d6456
active boot switches: -d:danger

[SolitudeSF]

nim's bundled cfg is evaluated before user/project one, so release and danger have no effect in them (compiler options are already set when user cfg is evaluated).

[disruptek]

The issue is that the compiler shouldn't report that it created a danger build when it hasn't.

[ghost]

Should this be closed and added to #14272 or something?

[disruptek]

Perhaps we should wait until the problem is resolved to close the issue.