Open Robadob opened 2 days ago
Tbh the actual test looks a bit janky, it should probably just be a test that runplan constructor doesn't throw an exception, this equivalence doesn't really match the new/delete behaviour of the C test.
Though arguably, I'd prefer swig to detect when a class of wrong type or None
is passed to equality and always return false. This same error could be present throughout the codebase. Can't find any obvious documentation for how SWIG should handle this edge-case, and not clear in the slightest how it ever didn't occur given I managed to reproduce it with SWIG 4.0.2.
Can confirm this error is occuring under linux too, with Python 3.10 and Swig 4.2.1, however a local build using Swig 4.0.2 passess these tests successfully, implying it might be a change in swig?
From users writing python, it should be possibel to compare to None
without haveing to also check the type first, so it is something we should try and resolve so that the test (which has been unchanged in 3 years since it was originally written) passes.
The rc1 wheels from our wheel house which were built with 4.0.2 also still pass the test.
Couple of Python test failures (debug/windows) whilst investigating the SWIG 4.2.1 issue, note this is with the Python wheel built with SWIG 4.2.1. Although, I've built it with 4.0.2 and received the same failures.
Not immediately clear their cause, these weren't touched in the recent bugfix so that's a coincidence.