sql: divide udf logic tests into separate files #106729
Conversation
|
This change is |
rharding6373
force-pushed
the
20230711_udf_logic_test_split_106471
branch
3 times, most recently
from
b87d7a7 to
3c2c089
Compare
|
You can hold off on reviewing until I've added some of the newly proposed logic test style to the new test files (i.e., use subtest end, include pg error codes). |
👍 Nice! But if that becomes onerous, don't feel obligated to clean all that up now. |
rharding6373
force-pushed
the
20230711_udf_logic_test_split_106471
branch
2 times, most recently
from
67f7640 to
6208df4
Compare
|
Ready for review now. I did not investigate the few questionable pg error codes yet, opting to leave them as a TODO for #107369. |
There was a problem hiding this comment.
Thanks for doing this! Looks like it was very tedious...
Reviewed 1 of 16 files at r1, 21 of 21 files at r2, 17 of 17 files at r3, all commit messages.
Reviewable status:complete! 1 of 0 LGTMs obtained (waiting on @rharding6373)
Break up the UDF logic test into multiple files based on functionality in order to reduce risk of timeouts in long-running configurations. Epic: none Fixes: cockroachdb#106471 Release note: none
Used postgres where possible to generate the error codes. A few questionable codes have been marked with TODOs to be addressed in future work. Epic: none Informs: cockroachdb#107369 Release note: none
rharding6373
force-pushed
the
20230711_udf_logic_test_split_106471
branch
from
6208df4 to
b959f49
Compare
|
TFTR! I wrote a script [0] to try to make parts of the logic tests runnable in postgres to try to get the expected error codes, which was somewhat helpful but far from 100%. Needs some tweaking before it could be used to get error codes for all logic tests. Or maybe there's an easier way to run logic tests against postgres. [0] https://gist.github.com/rharding6373/c6982084a2f810ebc8b754cd179a98bb |
|
bors r+ |
|
Timed out. |
|
Bors says that CI for this batch succeeded, but the overall state failed: https://bors.crdb.dev/batches/35879, which seems like a test infra problem. Going to try one more time, since it seems like the UDF logic tests themselves finished well under the timeout. bors r+ |
|
Build succeeded: |
Break up the UDF logic test into multiple files based on functionality in order to reduce risk of timeouts in long-running configurations.
Epic: none
Fixes: #106471
Release note: none