bpo-44991: Make GIL handling more explicit in sqlite3 callbacks

[erlend-aasland]

• acquire the GIL at the very start[1]
• release the GIL at the very end

[1] The trace callback performs a sanity check before acquiring the GIL

https://bugs.python.org/issue44991

Automerge-Triggered-By: GH:encukou

[erlend-aasland]

FYI, I'll normalise the callback names in a separate PR.

[encukou]

_destructor is also a callback function. It should get & release the GIL itself. (The NULL check can be moved here as well, as other calls of free_callback_context are definitely with a non-NULL context.)
PyGILState_Ensure/PyGILState_Release can be then removed from both create_callback_context and free_callback_context. Those will always be called with the GIL held.

[erlend-aasland]

_destructor is also a callback function. It should get & release the GIL itself. (The NULL check can be moved here as well, as other calls of free_callback_context are definitely with a non-NULL context.)
PyGILState_Ensure/PyGILState_Release can be then removed from both create_callback_context and free_callback_context. Those will always be called with the GIL held.

Yes, that's cleaner. Thanks! See d42eb38.

PTAL :)

[encukou]

Looks good, thank you!

[erlend-aasland]

Thanks, and thank you for reviewing!

[miss-islington]

@erlend-aasland: Status check is done, and it's a failure ❌ .

[miss-islington]

@erlend-aasland: Status check is done, and it's a success ✅ .