Add support for PyPy 3.10 #113
Conversation
sethmlarson
force-pushed
the
pypy-support
branch
2 times, most recently
from
f6e8faf
to
3f40b6c
Compare
|
Oh my. Mostly I'm just impressed if it works. Good call on only changing it for non-CPython implementations, in case someone is already using isinstance to check if they've got our SSLContext. I'll wait for final review until the tests pass. |
sethmlarson
force-pushed
the
pypy-support
branch
from
3f40b6c
to
03c47c5
Compare
|
Looks like a bug in PyPy's |
|
Okay @davisagli tests are passing so ready for review. If you'd like to take a stab at PyPy support without the horrible hacks please do, I would like to not have them 😓 |
| # Dirty hack to get around isinstance() checks | ||
| # for ssl.SSLContext instances in aiohttp/trustme | ||
| # when using non-CPython implementations. | ||
| return _truststore_SSLContext_dunder_class or SSLContext |
There was a problem hiding this comment.
Can you point me to a reproducer for why this was needed for PyPy? Is there something PyPy should be doing to make isinstance work?
| certfile=mkcert_certs.cert_file, | ||
| keyfile=mkcert_certs.key_file, | ||
| certfile=str(mkcert_certs.cert_file), | ||
| keyfile=str(mkcert_certs.key_file), |
There was a problem hiding this comment.
Sorry about that, is there a simple reproducer for this? PyPy passes the CPython ssl stdlib tests, maybe there is a case missing there?
|
Thanks for getting PyPy to work. I am not familiar with the project, but am a PyPy core dev. Please ping me (mattip) if you come across any other PyPy problems. |
|
I opened https://foss.heptapod.net/pypy/pypy/-/issues/4002 to continue the discussion. |
|
@mattip Thanks Matti! We'll keep you in mind if we run into future problems :) |
This is very much a hack, I don't like it, but attempting to add support to PyPy while maintaining the
SSLContextsubclassing is a huge pain from my initial attempt. The__class__hack maintains compatibility for aiohttp and trustme.Closes #112