The behavior of random.seed() with bytes and version=1 is specially documented. But actually it doesn't work.

>>> import random
>>> random.seed(b'abc', version=1)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/serhiy/py/cpython/Lib/random.py", line 113, in seed
    x = ord(a[0]) << 7 if a else 0
TypeError: ord() expected string of length 1, but int found

New changeset 132a7d7cdbc7cb89fa1c1f4e8192241c3d68f549 by Raymond Hettinger in branch 'master':
bpo-31482:  Missing bytes support for random.seed() version 1 (#3614)
https://github.com/python/cpython/commit/132a7d7cdbc7cb89fa1c1f4e8192241c3d68f549

Raymond, why the fix is not backported to 3.6?

Mariatta, what is the procedure for completing the backport?  I can see the miss-islington bot made a commit but it didn't take in the 3.6 branch.

What is wrong with PR 3629?

Hi Raymond, 

Somehow the backport PR https://github.com/python/cpython/pull/3629 was closed before instead of merged.

To retrigger the backport, we can re-apply the "needs backport to 3.6" label on the original PR (3614). I will do this now :)

Backport PR has been created. Once the CI checks are done, we can squash and merge it.

If PR 3629 was closed by mistake it could be just reopened, isn't?

Serhiy, it can't be re-opened since miss-islington deleted the branch already.

Ah, now I understand what happened. Thanks Mariatta.

New changeset c6ce8fba07ea6798eac46ab2808167afecd4d90b by Mariatta (Miss Islington (bot)) in branch '3.6':
[3.6] bpo-31482:  Missing bytes support for random.seed() version 1 (GH-3614) (GH-3659)
https://github.com/python/cpython/commit/c6ce8fba07ea6798eac46ab2808167afecd4d90b

Backport was done. Thanks.