Coroutines do not unwind the stack in case of errors

Back to how it was, a coroutine does not unwind its stack in case of
errors (and therefore do not close its to-be-closed variables). This
allows the stack to be examined after the error. The program can
use 'coroutine.kill' to close the variables.

The function created by 'coroutine.wrap', however, closes the
coroutine's variables in case of errors, as it is impossible to examine
the stack any way.

testes/db.lua

@@ -734,18 +734,24 @@ a, b = coroutine.resume(co, 100)
 assert(a and b == 30)
 
 
--- check traceback of suspended coroutines
+-- check traceback of suspended (or dead with error) coroutines
+
+function f(i)
+  if i == 0 then error(i)
+  else coroutine.yield(); f(i-1)
+  end
+end
 
-function f(i) coroutine.yield(i == 0); f(i - 1) end
 
 co = coroutine.create(function (x) f(x) end)
 a, b = coroutine.resume(co, 3)
 t = {"'coroutine.yield'", "'f'", "in function <"}
-repeat
+while coroutine.status(co) == "suspended" do
   checktraceback(co, t)
   a, b = coroutine.resume(co)
   table.insert(t, 2, "'f'")   -- one more recursive call to 'f'
-until b
+end
+t[1] = "'error'"
 checktraceback(co, t)