Dirk Mueller
51bce7483e
* Remove the random state caching, which would grow without
bound, leaking memory in long test runs. The caching was
added to slightly speed up re-using the same (final) seed,
but since the final seed is now different for each test, it
has no effect. PR #690.
* Modify Numpy seed restriction, replacing hashing with a
modulo operation. The extra work to hash is unnecessary now
that we generate a final seed per test with CRC32. This
change saves ~500ns per test when Numpy is installed. PR
#691.
* Support Python 3.14.
* Use a different random seed per test, based on the test ID.
This change should mean that tests exercise more random data
values in a given run, and that any randomly-generated
identifiers have a lower chance of collision when stored in a
shared resource like a database. PR #687. Thanks to Bryce
Drennan for the suggestion in Issue #600 and initial
implementation in PR #617.
* Move from MD5 to CRC32 for hashing test IDs, as it’s 5x
faster and we don’t need cryptographic security. Issue #686.
OBS-URL: https://build.opensuse.org/package/show/devel:languages:python:pytest/python-pytest-randomly?expand=0&rev=40
4 lines
130 BLFS
Plaintext
4 lines
130 BLFS
Plaintext
version https://git-lfs.github.com/spec/v1
|
|
oid sha256:dc0ad4fc1ff37344b37026063b9587596052215c670094439d68e0eab72cd684
|
|
size 52023
|