CI jobs have gone from ~6 minutes to timing out at 6 hours

[ryjones]

this happened after #111 was merged.

[M-AlNoaimi]

Hi, #111 was initially made to actually fix the failing CI that was ongoing in the main branch after #107 was merged in. Local testing (back in May with forcing the latest version of liboqs at that time) was able to fix the CI issues, the idea was to skip the incompatible tests via the PR: 2142 - which worked, but CI of the #111 kept failing because liboqs-python pulls a specific version of liboqs-c instead of the latest release.

I noticed the PR #119 which attempts at merging & updating a couple changes, #116 adds quite a bunch of changes that @Guiliano99 clarified over at @vsoftco's PR to how to get the CI working. I suggest maybe trying to test out liboqs-python commit 328ca53 + #111's proposed changes & updating the version of liboqs being installed by the CI to see if it works then addressing the changes done at #116.

[Guiliano99]

Hi, I noticed that the OQS typing was missing, which caused key deserialization to fail in the pipeline (locally it wasn’t needed). Should I fix this in two separate pull requests?

• For the tests: I updated the pipeline so it only runs the fast XMSS/XMSS^MT options and loads the slower height-20 (XMSS) keys, but I assume I should hold off on that for now? The simplified pipeline currently takes about 12 minutes.

• Regarding LMS: it’s not runnable at the moment because the keys can’t set a key storage. Additionally are some issues tracked in LMS Issues with Proposed Fixes #2229, which is why I deactivate those algorithms for the tests.

• Finally, I pinned the liboqs version since changes between 0.13.1-dev and 0.14 required updates to the StatefulSignature wrapper. Since it wasn’t pinned before, I thought it would be safer while working on this. Or was the idea to keep it unpinned so that updates would not be need as long as the code keeps working?

[dstebila]

Fixed by #126.