Ada policy

[Richard Kenner]

    Primary languages and platforms should have them, they come in handy
    from time to time and help make the compiler more maintainable and
    help ensure a certain quality standard.  Do no harm is really more
    interesting to a customer than a random bug fix that they don't hit,
    and by far, most code doesn't hit most bugs, therefore the greater
    good is to not fix the bug, if you can't also do the testcase to
    ensure that the customer won't hit the bug again...

The key point is who is "you".  ACT can clearly put the bug in *their*
regression suite, so the bug will stay fixed.  Not fixing a customer
bug isn't an option.

The point is what to do with the FSF tree with respect to fixes for
bugs for which no resources are available to make a non-proprietary
test case.  Should the fix for the bug not be propagated to the FSF
tree?  I see no argument in favor of doing that.