Fix "duplicate key value violates unique constraint" on needle updates #5426
Conversation
This helps on systems where another timezone is generally used (and thus also used by PostgreSQL) to get consistent behavior when running tests. Note that this is relevant for code that uses the SQL function `now()` instead of `DBIx::Class::Timestamps::now` which already always uses UTC anyway.
|
Of course it would be great to test this as suggested here: https://progress.opensuse.org/issues/153616#Suggestions However, I would have to find out how to bring this change into the test setup first. |
This is not trivial. |
* Use raw SQL queries with "ON CONFLICT DO NOTHING" to avoid running into
those errors
* Follow example from `OpenQA::Schema::ResultSet::Assets::register` where
we already use raw queries like that followed by an DBIx call to get the
newly inserted row
* Not using `{for => 'update'}` here to locking as we likely don't need it
here as we don't mind the *exact* last module
* Avoid repeated function calls simplifying the code a little bit
* See https://progress.opensuse.org/issues/153616
Martchus
force-pushed
the
needle-race-cond
branch
from
37580e2 to
bf3149e
Compare
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #5426 +/- ##
=======================================
Coverage 98.37% 98.37%
=======================================
Files 389 389
Lines 37724 37728 +4
=======================================
+ Hits 37112 37116 +4
Misses 612 612 ☔ View full report in Codecov by Sentry. |
|
Looks very good: https://openqa.opensuse.org/tests/overview?version=Tumbleweed&build=dheidler_mm3_ping&distri=openqa&groupid=38 The first 10 out of 10 all passed. |
those errors
OpenQA::Schema::ResultSet::Assets::registerwherewe already use raw queries like that followed by an DBIx call to get the
newly inserted row
{for => 'update'}here to locking as we likely don't need ithere as we don't mind the exact last module