Jukka Lehtosalo

Because of the issues outlined above, we should probably start with lazily deserializing cache files. This would be easier to implement, and it would have some other nice properties: *...

Lazily processing entire SCCs (sets of cyclically dependent modules) may still be practical, even when they are not cached. We'd always process a third-party SCC imported from first-party code in...

I've thought more about this, and there's another interesting approach that could work. I'm not sure which of the options discussed here is the best yet, but this might be...

@ilevkivskyi recently contributed as new binary cache format, with fully custom deserialization and serialization code. Beyond massively improving deserialization speed, this seems to enable a simple implementation of partial lazy...

I've seen projects use stubs targeting an older version than the installed version of a library, since updating stubs can sometimes take significant effort. In particular, a new version of...

It seems sufficient to call the generated module init function to create and populate the module object, and then put the module object to `sys.modules` with the help of `PyImport_GetModuleDict()`....

One more option: 5. Migrate test cases to regular .py files from .test files, and require unique module names for test modules.

After some more thinking, I think option 5 is the best, even though it will take a bit of effort to implement. It would enable syntax highlighting and other IDE...

I think the malware scanner is triggered when a compiled binary is executed for the first time. I think it's a feature of recent macOS versions, and I haven't found...

The PR below helps with slow tests. `runtests.py mypyc-fast` is now about 50-100% faster that running the whole mypyc test suite: * python/mypy#17906 This issue is still relevant, but I...