Tuple is tricky. If there is a bug in tuple you can bet that SFINAE is both the 
problem and the solution. constexpr constructors are becoming a frequent 
problem in the tuple SFINAE because they tend to cause hard compile errors. The 
SFINAE you added could easily introduce an almost identical bug in a different 
scenario.

Another reason I'm so hesitant to move forward with this patch is because they 
essentially contradicts http://reviews.llvm.org/D8178.

The first step to solving both of these bugs is to drastically increase the 
test coverage of tuple. The tests need to cover a much wider range of types 
than they currently do. Until then I think both these patches will need to wait 
in limbo.


REPOSITORY
  rL LLVM

http://reviews.llvm.org/D9085

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/



_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to