Matthias Urlichs
Matthias Urlichs
.You've described in a lot of words why I didn't bother with sans-io-ing any of my code yet. There's two missing building blocks in sans-io. One is timeout handling, including...
On the receive side we do need a buffer, obviously; equally obviously that buffer needs to be part of the object's interface. In fact we need to be able to...
> We don't want to add buffer management to the Stream interface, because there are many different classes that implement that interface and we don't want them all to be...
IMHO `.pyi` files only make sense when we can't change the `.py` files, e.g. because they're shipped by somebody else. This doesn't apply to Trio. See also https://github.com/dropbox/pyannotate for a...
Sure, but we don't actually have any of these AFAIK.
I tend not to do that, as a connection pool works better as a stack: the data of the connection that's been given back most recently are still in the...
Umm, thou shalt not swallow the `Cancelled` exception (in fact I'd ignore the internals and catch+reraise `BaseException` instead), but yes, that's a reasonable use case.
Meh. Frankly I don't think we can achieve 100% backwards compatibility, which is especially bad since the affected code is on code paths that are typically not tested very well....
I have no idea why these test failures happen, but they're very unlikely to be from my change.
trio-asyncio just uses a contextvar to store the loop. That works for applications because you typically have one loop at top level and there is no "outside" to call into...