Brian McFee

> Do you see anything suggesting otherwise? Yes, your response to the bug report form: > ### software versions > > ``` > mitmanek:python-librosa (1A1M) $ python3 > Python 3.11.10...

> OK, once more. We are building packages in openSUSE so that there are multiple versions of binary packages generated, currently one for Python 3.10, one for 3.11, one for...

Definitely worth fixing! I'd second @tkoeppe 's suggestion of dropping the sphinx/reST cruft in favor of unicode. That doesn't necessarily help with the `\*` or `\%`, but I think that's...

> @bmcfee can you take a look at the requested changes and and then I can merge? I may be having some kind of brain meltdown here, but I don't...

Note: version bump here would at minimum be to 1.8, which is fine for us according to SPEC0 #1935 . Probably we can get away with a higher minimum version...

Can you run this through cProfile (or using the `%profile` magic in ipython/jupyter) so we can get an accurate time measurement for everything that happens? Your code example is not...

@johan010 ok, my response above applies in your case as well.

Thanks for the additional info. The main difference between 0.9.2 and 0.10.x would be lazy-loading. (Unless your downgrade also triggered dependency version downgrades - this is why a full environment...

> I think this behavior is probably not a bug, but a feature to improve call times of the functions, however it does come as a disadvantage in some use...

I did a bit more digging and profiling of imports, and I'm not seeing a clear smoking gun here. There are some things that we could probably defer to speed...