Bug 6967 – template instantiation depends an pragma(msg, T.stringof) if __traits is used

From a thread in d.learn: http://www.digitalmars.com/d/archives/digitalmars/D/learn/Strange_behaviour_in_templates_Another_bug_30591.html I'm trying to learn template meta programming in D. To do this, I'm working on a  meta state machine like Boost.MSM  (http://www.boost.org/doc/libs/1_47_0/libs/msm/doc/HTML/index.html). States are types and can have sub states. If they do, they need to have an attribut 'SubStates', which should be a TypeTuble with the  sub states. The task at hand is: Given a State and a list of all States, create a list of all parents of State including State itself. This code (http://paste.pocoo.org/show/506154/) ought to do this and it works sometimes.  You can make to interesting observations of this code: In its current form, 'dmd test.d -unittest' will print: " > false But if you uncomment the first pragma() in the unittest block, it will print  > ()  > true If you change the body of ParentsDesc to  > alias TypeTuple!() Result; it will also print > true Although the static if in ParentsDesc is never true in this example.

The code (http://paste.pocoo.org/show/506154/) had disappears, so we cannot know what was the original issue.