Returning by-ref values from overridden methods and not capturing them at call site

[e-kayrakli]

Summary of Problem

Returning some values from overridden methods seems to cause compiler crashes, if the value returned is not captured at the call site.

(I am not entirely sure if the title of the issue captures the problem in a good way)

A little bit of GDB'ing shows that calledFn in https://github.com/chapel-lang/chapel/blob/master/compiler/resolution/callDestructors.cpp#L515 seems to be null if the call is a PRIM_VIRTUAL_METHOD_CALL causing a null dereference in the next line.

Steps to Reproduce

Source Code:

class A {
  proc foo() { return (3, 3); }
}

class B:A {
  proc foo() { return (1,1); }
}

var b:A = new B();
b.foo();

I have tried and confirmed that this crashes if foo returns: homogeneous/heterogeneous tuples and strings. Returning a class instance field of A seems to compile fine, any primitives are also fine. Moreover returning a primitive class field as such also seems fine:

proc foo() ref { return someIntField; }

Compile command:
chpl foo.chpl

Execution command:
Doesn't compile.

Associated Future Test(s):
test/classes/elliot/non-captured-virtual-return.chpl #7318

Configuration Information

• Output of chpl --version:

chpl Version 1.16.0 pre-release (405f2a2a7a)
Copyright (c) 2004-2017, Cray Inc.  (See LICENSE file for more details)

• Output of $CHPL_HOME/util/printchplenv --anonymize:

CHPL_TARGET_PLATFORM: linux64
CHPL_TARGET_COMPILER: gnu
CHPL_TARGET_ARCH: native
CHPL_LOCALE_MODEL: flat
CHPL_COMM: none
CHPL_TASKS: qthreads
CHPL_LAUNCHER: none
CHPL_TIMERS: generic
CHPL_UNWIND: none
CHPL_MEM: jemalloc
CHPL_MAKE: make
CHPL_ATOMICS: intrinsics
CHPL_GMP: gmp
CHPL_HWLOC: hwloc
CHPL_REGEXP: re2
CHPL_WIDE_POINTERS: struct
CHPL_AUX_FILESYS: none

[e-kayrakli]

P.S this was also something discovered by @LouisJenkinsCS

[mppf]

@e-kayrakli , @LouisJenkinsCS - this bug should be fixed in master now. Could you check over the Collections modules & update any documentation / fix any workarounds? Once you do that we should close this issue. Thanks.

[LouisJenkinsCS]

I'll have it done soon.

[LouisJenkinsCS]

@e-kayrakli

You can close the issue now.