So far I've been moving patches with failing tests to Waiting on Author but there are a number with "minor" failures or failures which look unrelated to the patch. There are 20 patches with at least one failing test in Ready for Comitter (2) and Needs Review (18).
Here's a summary of the reasons for these failures. Mostly they're bitrot due to other patches but there are a few test failures and one where Cirrus is just serving up a blank page... Ready for Committer: 37/3496 Avoid erroring out when unable to remove or parse logical rewrite files to save ... [14:12:40.557] t/013_crash_restart.pl (Wstat: 256 Tests: 18 Failed: 1) [14:12:40.557] Failed test: 5 37/3574 Frontend error logging style Patch bitrot: plan is to let this sit till the end of the commitfest, then rebase and push Needs Review: 37/3374 Allows database-specific role memberships Patch has bitrotted. It does seem a bit unfair to ask authors to keep rebasing their patches and then not review them until they bitrot again .... I guess this is a downside of concentrating a lot of commits in the CF. 37/3506 CREATEROLE and role ownership hierarchies Thread digression was committed but main patch is still pending. cfbot is testing the digression. 37/3500 Collecting statistics about contents of JSONB columns Patch bitrotted, presumably due to the other JSON patchset. I moved it to Waiting on Author but I think it's mainly in need of design review so it could still be useful to look at it. 37/2433 Erase the distinctClause if the result is unique by definition Patch has bitrotted. But there are multiple versions of this patchset with different approaches and there is a lot of divergence of opinions on which way to go. I'm inclined to mark this "returned with feedback" and suggest to start with fresh threads on individual parts of this patch. 37/2266 Fix up partitionwise join on how equi-join conditions between the partition keys... [08:10:46.390] # poll_query_until timed out executing this query: [08:10:46.390] # SELECT '0/13AC5E80' <= replay_lsn AND state = 'streaming' [08:10:46.390] # FROM pg_catalog.pg_stat_replication [08:10:46.390] # WHERE application_name IN ('standby_1', 'walreceiver') 37/2957 Identify missing publications from publisher while create/alter subscription. It looks like this was committed, yay! Please remember to update the CF app when committing patches. It's sometimes hard for me to tell if it was only part of a patch or a related fix discovered in discussing a patch that was committed or if the commit resolves the entry. 37/2218 Implement INSERT SET syntax Tom: "This patch has been basically ignored for a full two years now (Remarkably, it's still passing in the cfbot.)" Well it has bitrotted in gram.y now :( 37/2138 Incremental Materialized View Maintenance bitrotted in trigger.c. This is a huge patchset and probably hard to keep rebased and I think still looking for more fundamental design review. 37/3071 Lazy JIT IR code generation to increase JIT speed with partitions [00:01:20.167] t/013_partition.pl (Wstat: 7424 Tests: 31 Failed: 0) [00:01:20.167] Non-zero exit status: 29 [00:01:20.167] Parse errors: No plan found in TAP output 37/3181 Map WAL segment files on PMEM as WAL buffers [19:38:40.880] configure: error: library 'libpmem' (version >= 1.5) is required for PMEM support Not sure if this is just the expected output if a platform lacks this library or if the tests need to be adjusted to add a configure option. 37/3052 Merging statistics from children instead of re-sampling everything Uh, Cirrus is just giving me a blank page on this one. No idea what's going on here. 37/3490 Pluggable toaster An updated patch has been posted since the failure 37/1712 Remove self join on a unique column Patch bitrotted in a regression test expected output 37/3433 Removing more vacuumlazy.c special cases, relfrozenxid optimizations [06:50:33.148] vacuum.c: In function ‘vac_update_relstats’: [06:50:33.148] vacuum.c:1481:6: error: ‘oldrelminmxid’ may be used uninitialized in this function [-Werror=maybe-uninitialized] [06:50:33.148] 1481 | errmsg_internal("overwrote invalid pg_class.relminmxid value %u with new value %u in table \"%s\"", [06:50:33.148] | ^~~~~~~~~~~~~~~ [06:50:33.148] vacuum.c:1475:6: error: ‘oldrelfrozenxid’ may be used uninitialized in this function [-Werror=maybe-uninitialized] [06:50:33.148] 1475 | errmsg_internal("overwrote invalid pg_class.relfrozenxid value %u with new value %u in table \"%s\"", [06:50:33.148] | ^~~~~~~~~~~~~~~ 37/2901 SQL/JSON: functions Patchset is mostly applied so the cfbot is seeing patch conflicts. 37/3589 Shared memory based stats collector cfbot is trying to apply a digression patch rather than the main patch (which is up to version 67! and has 17 parts) 37/3546 Support custom authentication methods using hooks Patch bitrot in hba.c. But the discussion is mainly about the design and whether it's "the right way to go" anyways. 37/3048 pg_stat_statements: Track statement entry timestamp [13:19:51.544] pg_stat_statements.c: In function ‘entry_reset’: [13:19:51.544] pg_stat_statements.c:2598:32: error: ‘minmax_stats_reset’ may be used uninitialized in this function [-Werror=maybe-uninitialized] [13:19:51.544] 2598 | entry->minmax_stats_since = minmax_stats_reset; [13:19:51.544] | ~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~ -- greg