Fix flaky postgres test #7228
Fix flaky postgres test #7228
Conversation
Codecov Report
@@ Coverage Diff @@
## master #7228 +/- ##
==========================================
- Coverage 94.03% 94.01% -0.03%
==========================================
Files 172 172
Lines 12913 12913
==========================================
- Hits 12143 12140 -3
- Misses 770 773 +3
Continue to review full report at Codecov.
|
|
Do you have to use an extension for this? |
Trying to see if I can install the extension directly for Postgres 10 and 11, seems to already be included in 13 (and I think 12) |
|
I just realized that the Contributing Guide need to be updated for testing against Postgres like creating database and proper extensions. |
|
The contributing guide shows how to add the correct extensions for regular usecases |
|
Which line? |
137 to 139 I always copy paste from the Contributing guide when testing to make sure it's still valid |
|
I see, I was looking at the old travis reference that is in there. |
I can update this here since I'm adding the pgcrypto extension |
|
@dplewis test look all good. I'm assuming I don't need to add a changelog entry for a test fix? |
|
🎉 This change has been released in version 5.0.0-beta.1 |
|
🎉 This change has been released in version 5.0.0 |
New Pull Request Checklist
Issue Description
I introduced a flaky test in #6506 where MD5 was used to create random strings. The problem here is these strings are used for values on a unique index and MD5 generates clashes easily as the strings are not promised to be unique.
Related issue: Flaky tests
Approach
Use postgres
gen_random_uuid()instead of MD5 for random strings since they are guaranteed to be unique, especially for the amount generated in the tests (5000).TODOs before merging