[3.6] bpo-32947: Fixes for TLS 1.3 and OpenSSL 1.1.1#11612
[3.6] bpo-32947: Fixes for TLS 1.3 and OpenSSL 1.1.1#11612ned-deily merged 1 commit intopython:3.6from vstinner:test_ssl36
Conversation
Backport partially commit 529525f: complete the previous partial backport (commit 2a4ee8a. Co-Authored-By: Christian Heimes <[email protected]>
|
@ned-deily: I don't know your policy for tests which fail randomly. This one is impacting RHEL builder. We will backport it in RHEL, but I don't know if it's worth it to backport is also in 3.6 upstream? Do we care to have a highly reliable test suite for 3.6? It's not like we push changes frequently in 3.6 anymore. To be clear: I don't really care if you reject this change, it's up to you :-) |
|
I'm OK with backporting this to 3.6 to ensure that the tests keep working as long as @tiran is OK with it. Awaiting his review. |
|
@tiran: Would you mind to review this change? Right now, "./python -m test -u all -F -m test_connect_cadata test_ssl" fails on Fedora 29 with OpenSSL 1.1.1 FIPS. I tried backported 2 changes from your master change, 2 changes that you forgot and chose to not backport in your backport to 3.6. See the PR description and https://bugs.python.org/issue32947#msg333990 |
|
test_ssl failed randomly on the Fedora CI: I decided to backport this test_ssl fix into python36 package of Fedora:
cc @stratakis |
|
@stratakis backported most OpenSSL 1.1.1 and TLS v1.3 changes from 3.6 to 2.7, but we didn't backport this change since https://bugs.python.org/issue32947 doesn't affect Python 2.7. It's maybe because test_ssl handles exceptions differently: exception hierarchy in Python 2 is very different than Python 3 exception hierarchy. |
|
@tiran: Would you mind to review this change? |
|
@ned-deily: Would you mind to review/merge this PR? |
|
Backport partially commit 529525f:
complete the previous partial backport (commit
2a4ee8a.
Co-Authored-By: Christian Heimes [email protected]
https://bugs.python.org/issue32947