this repo has no description
evaluator: add regression test to lock in evalv3 wrt #3567
With evalv3, we correctly get an error about the nested call to
strings.Join missing a second argument. With evalv2, the error is
swallowed and we incorrectly get a success exit code and confusingly
zero output.
Fixes #3567.
Signed-off-by: Paul Jolly <paul@myitcv.io>
Change-Id: I12a6e7c31b829800dfce839c4e0f193f32ec38a9
Reviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1205366
TryBot-Result: CUEcueckoo <cueckoo@cuelang.org>
Unity-Result: CUE porcuepine <cue.porcuepine@gmail.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
+66
+66
cue/testdata/builtins/issue_3567.txtar
+66
cue/testdata/builtins/issue_3567.txtar
···
1
+
-- in.cue --
2
+
import "strings"
3
+
a: strings.Join( [ strings.Join(["1"]), "2" ], "\n")
4
+
5
+
-- out/compile --
6
+
--- in.cue
7
+
{
8
+
a: 〈import;strings〉.Join([
9
+
〈import;strings〉.Join([
10
+
"1",
11
+
]),
12
+
"2",
13
+
], "\n")
14
+
}
15
+
-- out/eval/stats --
16
+
Leaks: 1
17
+
Freed: 5
18
+
Reused: 1
19
+
Allocs: 5
20
+
Retain: 2
21
+
22
+
Unifications: 6
23
+
Conjuncts: 6
24
+
Disjuncts: 7
25
+
-- out/evalalpha --
26
+
Errors:
27
+
0: not enough arguments in call to strings.Join (have 1, want 2):
28
+
./in.cue:2:20
29
+
30
+
Result:
31
+
(_|_){
32
+
// [eval]
33
+
a: (_|_){
34
+
// [eval] 0: not enough arguments in call to strings.Join (have 1, want 2):
35
+
// ./in.cue:2:20
36
+
}
37
+
}
38
+
-- diff/-out/evalalpha<==>+out/eval --
39
+
diff old new
40
+
--- old
41
+
+++ new
42
+
@@ -1,6 +1,12 @@
43
+
+Errors:
44
+
+0: not enough arguments in call to strings.Join (have 1, want 2):
45
+
+ ./in.cue:2:20
46
+
+
47
+
+Result:
48
+
(_|_){
49
+
// [eval]
50
+
a: (_|_){
51
+
- // [eval]
52
+
+ // [eval] 0: not enough arguments in call to strings.Join (have 1, want 2):
53
+
+ // ./in.cue:2:20
54
+
}
55
+
}
56
+
-- diff/explanation --
57
+
The old evaluator does not correctly handle the nested nature of calls
58
+
to strings.Join, and buries the error of the missing argument in the
59
+
nested call. The new evaluator handles this properly.
60
+
-- out/eval --
61
+
(_|_){
62
+
// [eval]
63
+
a: (_|_){
64
+
// [eval]
65
+
}
66
+
}