I know @JeffBezanson hates umbrella issues, but there are so many range-related problems cropping up that I wanted to have a single place to consolidate the various issues. Here are some of the relevant issues:

• ranges terribly sensitive to floating point issues – ranges terribly sensitive to floating point #2333
• integer range overflows – Integer range overflows #3079 and RFC: Basic BigRange support available #3087
• problem with rand(typemin(Int32):typemax(Int32)) – Problem with rand(typemin(Int32):typemax(Int32)) #5550 and Fix overflow in #5550 #5555
• performance / special case lowering – 0860767 and regression in performance of counting loops #5469

Less immediate, but related:

• first and last on empty ranges – first() and last() on empty ranges #2537
• Uint8 ranges – make [a, b] not concatenate #3737 (edit: more about cat than ranges)
• float ranges: mismatch between length and bounds error – float ranges: mismatch between length and bounds error #329 (closed)
• fix for last(t::Range) Fix for last{T}(t::Range{T}) #2734 (closed)
• fix float range behavior – Fix float range behavior #2959 (closed)
• repair floating point ranges with start==stop Repair floating point ranges with start==stop. #3961 (closed)
• infinite loop for Uint8 range – Infinite loop for Uint8 range #5483 (closed)

We need a more coherent story for ranges. I suspect that ranges of ordinal types, including integers and chars need to be handled rather differently from floating-point ranges and such.