Pablo Galindo Salgado
Pablo Galindo Salgado
> This is a breaking change right? Breaking how? This doesn't affect any visible thing that's guaranteed unless I am missing something. > We should also be more careful about...
I have added a translation layer similarly as we do for the REPL so it looks a bit better in tracebacks
> I don't think "backward-compatibility" is only about the things we "guarantee", it's also about the behavior that has been there for a long time, even without documentation. I understand,...
@lysnikolaou thoughts?
The problem is that it's not just the traceback: it's the code object. Most test here fail because either pdb or tracemalloc are getting the wrong name because of the...
> In linecache, when the filename starts with
> This will add some overhead when we register the code with `linecache`, but the majority of dynamically generated code is small and I don't think it'll make a big...
Ok, I think I misunderstood your proposal because I clearly had the wrong idea of how it would work. But then, what's the difference between this and having an "alternate"...
> If we have a reliable source for dynamically generated code, why keep it a secret? Because that's a new feature and we are considering a bugfix. (I'm on my...
> 3\. When we register code, put it in a different container than `linecache.cache` with the id of the code object as the key(different source code can build to the...