dmd segfaults on endless mixin recursion

[dlangBugzillaToGithub]

Dennis (@dkorpel) reported this on 2018-12-04T14:14:03Z

Transferred from https://issues.dlang.org/show_bug.cgi?id=19465

CC List

• Stefan Koch

Description

Code:
```
enum string s = "__traits(compiles, mixin(s))";

static if (__traits(compiles, mixin(s))) {
    enum b = 2;
}
```

Output:
Segmentation fault (core dumped)

Concocted example, not something I encountered in actual code. Still, "dmd should never segfault" so filing this as a minor bug.

[dlangBugzillaToGithub]

uplink.coder commented on 2018-12-04T16:43:06Z

Hmm, I think if you are asking for a stack-overflow you should not be surprised if you get it.

This is one of these problems which you get by interpreting a turing-complete language at compile-time.

Introducing an artificial recursion limit has been done, but people (weka) had to increase it in their fork.

Checking for a limit is not _huge_ deal but it is work which does not have to be done.

[dlangBugzillaToGithub]

dkorpel commented on 2018-12-04T21:42:58Z

Yeah, I don't know if/how this should be "fixed", but notably endless recursion in CTFE:
`` 
enum a = f();
int f() {return f();}
```
Results in:
Error: function `onlineapp.f` CTFE recursion limit exceeded

While endless recursion in mixin without __traits(compiles, ...):
```
enum string s = "mixin(s);";
mixin(s);
```
Results in, after a while:
Error: out of memory