[
https://issues.apache.org/jira/browse/IMPALA-14985?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18082048#comment-18082048
]
ASF subversion and git services commented on IMPALA-14985:
----------------------------------------------------------
Commit dc1321f4a6fa5acfb24d5db5bd8a02c2abc3f346 in impala's branch
refs/heads/master from Csaba Ringhofer
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=dc1321f4a ]
IMPALA-14985: Cleanup tests with unused 'vector' in EE tests
Having 'vector' argument leads to running a test as parametrized,
so running it with multiple parameter vectors. These parameters
don't affect how queries run unless passed to functions like
run_test_case() explicitly.
Note that unused 'vector's can give the false impression of
affecting the query options of queries run by default clients
as query option CLIENT_IDENTIFIER is actually modifed based
on the parameter vector using pytest hook pytest_runtest_logstart
and connections' clear_configuration and create_connection functions.
One example (TestIOMetrics) was found where the intention was to
pass num_nodes=1 but this didn't happen.
A test bug found during this change is that there were test classes
where "skip if not exhaustive" logic was implemented within
add_test_dimensions(), but this didn't affect non-parametrized
tests that were still run in core builds. This is fixed by
checking exploration strategy elsewhere.
The change also cleans up add_test_dimensions() functions which
are not used as the test class has no test function with
'vector' parameter.
Testing:
- passed exhaustive tests
Generated by: copilot + Claude Sonnet 4.6
Change-Id: I8e2a8fe333e33a906178a0821eb554890be26132
Reviewed-on: http://gerrit.cloudera.org:8080/24296
Reviewed-by: Noemi Pap-Takacs <[email protected]>
Tested-by: Csaba Ringhofer <[email protected]>
> Cleanup tests with unused 'vector' in EE tests
> ----------------------------------------------
>
> Key: IMPALA-14985
> URL: https://issues.apache.org/jira/browse/IMPALA-14985
> Project: IMPALA
> Issue Type: Improvement
> Components: Test
> Reporter: Csaba Ringhofer
> Priority: Major
>
> example:
> https://github.com/apache/impala/blob/53f5d74b9c905aad36c54f9251e0d25fc21d80bc/tests/catalog_service/test_large_num_partitions.py#L36
> test_list_partitions() has a vector argument, leading to potentially running
> it with multiple parameters, but actually the vector is not used so my
> understanding is that the all test runs will work the same
> the effect of vector can be misleading - for example the query options are
> not passed to default client, but CLIENT_IDENTIFIER query option is actually
> modified with a string containing all parameters in the vector, suggesting
> different query options (like disable_codegen)
> example:
> {code}
> Query Options (set by configuration):
> TIMEZONE=Europe/Budapest,CLIENT_IDENTIFIER=query_test/test_avro_schema_resolution.py::TestAvroSchemaResolution::test_avro_c_lib_unicode_nulls["table_format:avro/snap/block"|"protocol:hs2"|"exec_option:{'abort_on_error':1;'batch_size':0;'disable_codegen':True;'disable_codegen_rows_threshold':0;'
> {code}
> disable_codegen:true suggest no codegen, but actually a the query
> TestAvroSchemaResolution::test_avro_c_lib_unicode_nulls runs with codegen
> enabled
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
