fix: out of memory error when importing under postgresql If you try importing more than 20k items under postgresql you can run out of memory: psycopg2.errors.OutOfMemory: out of shared memory HINT: You might need to increase max_locks_per_transaction. Tuning memory may help, it's unknown at this point. This checkin forces a commit to the postgres database after 10,000 rows have been added. This clears out the savepoints for each row and starts a new transaction. back_postgresql.py: Implement commit mechanism in checkpoint_data(). Add two class level attributes for tracking the number of savepoints and the limit when the commit should happen. roundup_admin.py: implement pragma and dynamically create the config item RDBMS_SAVEPOINT_LIMIT used by checkpoint_data. Also fixed formatting of descriptions when using pragma list in verbose mode. admin_guide.txt, upgrading.txt: Document change and use of pragma savepoint_limit in roundup-admin for changing the default of 10,000. test/db_test_base.py: add some more asserts. In existing testAdminImportExport, set the savepoint limit to 5 to test setting method and so that the commit code will be run by existing tests. This provides coverage, but does not actually test that the commit is done every 5 savepoints 8-(. The verification of every 5 savepoints was done manually using a pdb breakpoint just before the commit. acknowledgements.txt: Added 2.4.0 section mentioning Norbert as he has done a ton of testing with much larger datasets than I can test with.

--- a/CHANGES.txt	Thu Oct 19 14:07:56 2023 -0400
+++ b/CHANGES.txt	Thu Oct 19 16:11:25 2023 -0400
@@ -59,6 +59,9 @@
 - Fix error handling so failure during import of a non-user item
   doesn't cause a second traceback. (Found by Norbert Schlemmer, fix
   John Rouillard)
+- Handle out of memory error when importing large trackers in
+  PostgreSQL. (Found by Norbert Schlemmer, extensive testing by
+  Norbert, fix John Rouillard)
 
 Features: