Merge v1.2.2 into main #17037
Merge v1.2.2 into main #17037
Conversation
* Only Finalize when all threads have been merged. fixes: duckdb#16617 fixes: dudblabs/duckdb-internal#4425
…into ext-bump-v121
…ity with core duckdb NaN ordering
I picked up the previous work on duckdb#16331 by @Tishj and added a test. As far as I can tell, setting the `gstate` to uninitialised is fine, here, and flushing before deleting fixes the issue.
This PR fixes duckdb#16942 As mentioned [here](https://duckdb.org/docs/stable/sql/dialect/sql_quirks#nan-values) DuckDB does not follow the IEEE-754 behavior for nan comparisons, instead it makes NaN the biggest value in the floating point ordering. This PR adjusts the pyarrow dataset filters we create when pushing DuckDB table filters in to pyarrow.
When embedding DuckDB, it is some times important to make sure all output follows a certain structure (example JSON) and/or goes to specific stream. This PR allows overriding the Printer's default behavior.
Co-authored-by: Carlo Piovesan <piovesan.carlo@gmail.com>
bump duckdb-httpfs, and duckdb-aws to latest version of v1.2-histrionicus Also removing `DONT_LINK` from avro since duckdb-iceberg tries to automatically load it, and if it is not available duckdb-iceberg will through an error. This means any time iceberg is autoloaded and avro is not available/linked, then statements/tests will fail. EDIT: not bumping iceberg in this PR. Want to try and isolate errors. EDIT2: Not bumping aws, there seems to be an issue with windows and provider `credential_chain`. Want to try and debug it on the windows laptop first
|
Sure, I can split this in 2, extensions vs all the rest. I will have a look in a bit. |
|
@Mytherin: I think some changes were still tied to extensions in v1.2-histrionicus, and this looks to me a reasonable compromise: bumping what can be bumping, and skipping in some places. Note that If Windows Extensions job passes this is good to go for me. |
|
Windows Extensions fails on what looks to me an irrelevant spatial test, unsure if path is skipping also spatial tests. |
|
Spatial test is fixed on spatial main branch - solution is just bumping to use that commit |
|
4be6065edc313a53ff2196ff79c11a0d5e249720 |
3316e73 to
9bf0866
Compare
9bf0866 to
7997d80
Compare
|
Thanks @Maxxen, moving to |
|
Thanks! |
Merge v1.2.2 into main (duckdb/duckdb#17037)
Merge v1.2.2 into main (duckdb/duckdb#17037)
Merge v1.2.2 into main (duckdb/duckdb#17037)
Merge v1.2.2 into main (duckdb/duckdb#17037)
Merge v1.2.2 into main (duckdb/duckdb#17037)
This brings in the latest changes in
v1.2-histrionicusintomain.The more complex bit was merging
.github/config/out_of_tree.cmake, that is still likely will not really compile due to patches being applied to different commits.Four regressions:
v1.2.2, rolling back a few commits to85ac4667bcb0d868199e156f8dd918b0278db7b9due to changes in Secret interfaceavrois currently disabled, needs a patch and re-enablingdeltais currently disabled, needs a different set of patches and re-enablingawstests are currently skippedAnd an unrelated change, that was needed due to incompatibility:
There are relevant issues to solve them and can be handled piece by piece.