|
From: | Ray Daniel Zimmerman |
Subject: | subsasgn / subsref bug |
Date: | Wed, 16 Jun 2021 22:31:27 +0000 |
I’ve encountered what I believe is a bug in subsref and
subsasgn handling in Octave 6.2, and would like some suggestions for how I might approach narrowing things down so I can even submit a useful bug report.
The issue is this … accessing obj.Properties inside methods of a classdef object with a property named “Properties” should not call the class’s
subsref method, but it sometimes does. And by “sometimes” I mean that it depends on something in the state of the Octave workspace or interpreter, not the particular code being executed.
Specifically, I have a rather large Matlab/Octave project with an extensive test suite and have recently begun to use both the
table class from Tablicious (thank you Andrew), if it is available, or my own very basic
mp_table class substitute, if it is not. Both
table and mp_table define subsref and
subsasgn methods, and both make assignments to fields of
obj.Properties within their constructors.
I got all my tests passing with both classes, but noticed that one particular test was failing with a max recursion limit error in
subsref when constructing a particular table (both with Andrew’s
table and my mp_table), but only if I ran certain other tests before running the one in which the failure occurs.
Just to clarify, there is no difference between the failing case and the working case in terms of the inputs supplied to the constructor in which this recursion error occurs. It appears to be something within the Octave interpreter that is getting
into a state in which it forgets that obj.Properties should not call
subsref.
Any suggestions for how I might narrow down what may causing this difference?
Thanks,
Ray
|
[Prev in Thread] | Current Thread | [Next in Thread] |