Conversation
fccffb0 to
d10cf01
Compare
|
Thank you for the pr, @nialov I don't like to have too much stuff in the conftest, so I would rather not move the proxy Could you please revert the proxy change, and instead create a new Thank you very much! |
|
Sure, will do! At the latest at the end of the week. |
d10cf01 to
57ce3b3
Compare
|
Done and hopefully it is according to your request! |
|
Thank you, what I did is part of what I wanted, but what I meant was to move everything gc-related from conftest to the module you just created, including the fixtures that were already there: pretty much this block and all the imports it requires, and clean up the unneeded imports from conftest.py: Lines 106 to 170 in 4747cdb |
57ce3b3 to
4e53c6b
Compare
All tests that have the 'gc' fixture are now marked with the refcount mark. The reasoning is that they demonstrate flaky behaviour and disabling them in certain CI is necessary to ensure reliable testing. See psycopg#692
4e53c6b to
40aacab
Compare
|
I hope it is one step closer again :D |
|
That looks good! Merging. I will also rebase it on the branch 3.1 and push there, so you can build the 3.1.x |
|
Thanks a lot! |
|
Done: pushed your changeset as 70ef364 on the maint-3.1 branch. If you want to build 3.1.14 you will have to patch the codebase yourself; from 3.1.15 on the changeset will be in place already. |
|
Yep, got it. Fetching the patch from there! |
All tests that have the 'gc' fixture are now marked with the refcount mark. The reasoning is that they demonstrate flaky behaviour and disabling them in certain CI is necessary to ensure reliable testing. See #692
Besides the suggested fixes by @dvarrazzo I moved come code from
fix_proxy.pytoconftest.pyas I believe now that other kinds of tests are also marked inpytest_collection_modifyitemsit probably does not belong in a file namedfix_proxy. Let me know what you think!