The PIC tests a bunch of conditions to see if it's ok to cache a property access by generating a stub or patching the inline path. I took these conditions from the property cache code. But there are a lot of conditions, so I might not have gotten it right. It should be checked again, maybe by comparing to what the tracer tests in test_property_cache. It would probably be ideal to better encapsulate cacheability testing. We could have an API that can be called right after doing a property get or set that would return whether the property is cacheable, and a structure describing the access path and all later validation that is required. This API could then be used independently by the interpreter, methodjit, and tracejit.
Status: NEW → RESOLVED
Last Resolved: 5 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.