Testing & Tooling #1587
Replies: 2 comments 2 replies
-
At least these don't cause spurious failures, do they? And they could all be ran easily with But I can definitely say that the build of odin-python is quite convoluted. Running tox takes ages, if you only want to run parts to cover your changes, you have to know what these are first, and the command can become quite long. Also, the build does not work on Windows, and AFAIK this is not documented anywhere. |
Beta Was this translation helpful? Give feedback.
-
Not addressing the bigger issue surrounding the builds, but the issue around the docs build was specifically triggered by tox cache causing issues w/ the docs build when there are changes in the api/sdk. I've updated the docs target in tox to always recreate its tox environment. https://github.com/open-telemetry/opentelemetry-python/pull/1560/files#diff-ef2cef9f88b4fe09ca3082140e67f5ad34fb65fb6e228f119d3812261ae51449R174 We should consider better strategies around tox caching where possible and maybe make the performance tests run separately in a job that is non blocking to the build. Performance testing on shared infrastructure is never very reliable... |
Beta Was this translation helpful? Give feedback.
-
It's lot of frustrating to deal with out of the blue CI failures. Some times docs build fails for no reason; docker-tests that run successfully all the time suddenly fails; pypy3 tests fails occasionally; benchmark-tests failures show up once in a while even when the changes aren't related. And there are multiple linters & formatters.
Please share your observations and thoughts. Hopefully we will have consolidated action item to work on post rc1.
Beta Was this translation helpful? Give feedback.
All reactions