Update the test for issue 211 to restore the old test behavior.

Also, test for looping computations in addition to `error` computations.
This doesn't let us cheat by eta-expanding the error, as we were
doing before.

tests/issues/issue211.icry

@@ -1,4 +1,8 @@
-take`{1} (join [0x1, 0x2, repeat (error "boom"), 0x4])
+take`{1} (join [0x1, 0x2, error "boom", 0x4])
+take`{1} (join [0x1, 0x2, x where x = x, 0x4])
 take`{1} [True, False, error "boom", True]
+take`{1} [True, False, x where x = x, True]
 (splitAt`{0} [True, False, error "boom", True]).0
+(splitAt`{0} [True, False, x where x = x, True]).0
 (splitAt`{0} [True, False, error "boom", True]).1
+(splitAt`{0} [True, False, x where x = x, True]).1

tests/issues/issue211.icry.stdout

@@ -1,9 +1,20 @@
 Loading module Cryptol
 0x0
+0x0
+0x1
 0x1
 0x0
+0x0
 
 Run-time error: boom
 -- Backtrace --
-Cryptol::error called at issue211.icry:4:28--4:33
-Cryptol::splitAt called at issue211.icry:4:2--4:9
+Cryptol::error called at issue211.icry:7:28--7:33
+Cryptol::splitAt called at issue211.icry:7:2--7:9
+
+<<loop>> while evaluating <interactive>::x
+This usually occurs due to an improper recursive definition,
+but may also result from retrying a previously interrupted
+computation (e.g., after CTRL^C). In that case, you may need to
+`:reload` the current module to reset to a good state.
+-- Backtrace --
+Cryptol::splitAt called at issue211.icry:8:2--8:9