I imagine this problem would have arisen using the old ARRAY.is_equal_map, but neither I nor the compiler ever used it.
Now that feature is called is_equal, it gets used and this is what happens.

THING contains an ARRAY[THING] and redefines is_equal to compare these arrays using is_equal. This is fine when two THINGs that are is_equal are compared, but when two not is_equal things are compared it gets into an endless recursive loop in which Current and other swap places in each alternate call.

Attached are the source-code files and a cut-down version of the traceback from the debugger after the crash.


Here's what happens when it runs:

ceres/2nd-try: compile -no_gc -g -no_strip test2_thing test -o test2
ceres/2nd-try: ./test2
thing_c fast_is_equal to thing_b
thing_c is_equal to thing_b
NOT thing_a fast_is_equal to thing_b
Segmentation fault
ceres/2nd-try:

In the gdb backtrace the symbols are:

r76  THING
r201 ARRAY[THING]
r202 NATIVE_ARRAY[THING]