| Age | Commit message (Collapse) | Author |
|---|
|
There are two issues at play here:
1. Python version of `map_hash` unnecessarily performs hash truncation
even if the hash is already 32-bit wide, which potentially converts
it from signed int to unsigned long.
2. The `test_none_collisions` test generates a collision node with
hash greater than 2^32.
Both of these are problematic on 32-bit systems, where `sizeof(Py_hash_t)`
is 4, and so anything that doesn't fit into `Py_hash_t` gets bit-mangled,
breaking the `hash(x) != x` invariance that the test relies upon.
Fixes: #53
Fixes: #50
|
|
|
|
|
|
* Improve typing of `__init__()`
* Update typing of `Map`-producing functions to produce the correct type
* Update typing of other methods to more closely align with `Mapping`
* Add protocol classes for unexposed data structures
* Export protocol classes for ease of use in typed code
* Update stub file to pass in mypy strict mode
|
|
Refs actions/runner#781, add LD_LIBRARY_PATH manually here for now
before an upstream fix is added.
|
|
* Fix flake8 errors and add it to CI
* Use extras for test dependencies
Co-authored-by: Fantix King <fantix.king@gmail.com>
|
|
|
|
|
|
|
|
|
