You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug] test_unit_gpversion tests the old Greenplum GpVersion behavior #2364
main (3.0.0-devel), commit c88c43d. REL_2_STABLE has the same test file.
What happened
gppylib.test.unit.test_unit_gpversion fails: 10 tests, 5 failures, 4 errors.
The test still describes the Greenplum GpVersion. The GpVersion in gpMgmt/bin/gppylib/gpversion.py has been different since the initial
Cloudberry code dump:
It accepts at most three version numbers. GpVersion([3,3,0,0]) fails with
"Version too long".
getVersionRelease() returns only the major number ("3", not "3.3").
isVersionCurrentRelease() compares only the major number with MAIN_VERSION, which is [3,99] on main. Old Greenplum strings in the test,
such as "3.3 build dev", are now the current release.
The << operator (__lshift__) is commented out, so test_lshift, test_50 and test_lshift_negative fail.
test_case_7 expects 7.0.0 to be the current release.
No CI job runs this module, so the failures were not seen.
What you think should happen instead
The test should check the current behavior: major-version releases, at most
three numbers, no <<, and the version strings that Cloudberry prints. It
should build its versions from MAIN_VERSION, so that it does not break at
the next major version.
This needs a rewrite of most of the test file, so I open this issue first.
Activity