On 9/18/16 8:14 AM, Andrei Alexandrescu wrote:
On 9/18/16 6:00 AM, Jonathan M Davis via Digitalmars-d wrote:
Yes. That's DIP 82:
http://wiki.dlang.org/DIP82
I need to go over it again and then introduce it into the new DIP
process.
But I really think that that's where we should go to fix this problem.
Just a thought: things that we can't do have high priority. Things that
we can do with a modest cost are much less attractive. Consider:
struct Awesome(A, B, C)
{
private enum ut = is(A == int) && is(B == int) && is(C == int);
static if (ut) unittest
{
...
}
}
You're looking at an overhead with a small fixed cost plus a few
characters ("if (ut)") per unittest.
This is exactly how I did RedBlackTree unit tests:
https://github.com/dlang/phobos/blob/master/std/container/rbtree.d#L748
https://github.com/dlang/phobos/blob/master/std/container/rbtree.d#L815
This is still less than ideal, and has caused some real problems over
the years. e.g.:
https://issues.dlang.org/show_bug.cgi?id=12246
https://issues.dlang.org/show_bug.cgi?id=14082
The best solution IMO is not to run templated unit tests unless
specifically requested. Perhaps only run them when instantiated inside a
unittest block?
All that being said, running unit tests on EVERY integral type caught
about 3 bugs in the compiler when I was creating dcollections, and found
several corner-case bugs in my code as well. It was well worth the
"overhead" IMO.
-Steve