While testing my own table-valued function, I got some rather weird results when running a window-function test. Try as I might, I couldn't account for it in my own code. The debugger showed the correct output being assigned on each call to the `xColumn()` function. So, I ran the same test with the `generate_series()` function as a sanity check, using the following query: CREATE TABLE persons(id INTEGER PRIMARY KEY, name TEXT); INSERT INTO persons (name) VALUES('John'), ('Paul'), ('George'), ('Ringo'); SELECT group_concat(value,name) OVER ( ORDER BY name ROWS BETWEEN 1 PRECEDING AND 1 FOLLOWING ) AS result FROM persons, generate_series(4450,4455); Granted, the query is nonsense, but the issue is not that the results don't make any sense; it's that information is missing: result ------------------------- 4450George4451 4450George4451George4452 4451George4452George4453 4452George4453George4454 4453George4454George4455 4454George4455John4450 4455John4450John4451 50John4451John4452 <-- initial values begin to be truncated 51John4452John4453 52John4453John4454 53John4454John4455 54John4455Paul4450 55Paul4450Paul4451 50Paul4451Paul4452 51Paul4452Paul4453 52Paul4453Paul4454 53Paul4454Paul4455 54Paul4455Ringo4450 55Ringo4450Ringo4451 450Ringo4451Ringo4452 451Ringo4452Ringo4453 452Ringo4453Ringo4454 453Ringo4454Ringo4455 454Ringo4455 shell `.version` command: SQLite 3.36.0 2021-06-18 18:36:39 5c9a6c06871cb9fe42814af9c039eb6da5427a6ec28f187af7ebfb62eafa66e5 zlib version 1.2.11 gcc-5.2.0 I compiled `series.c` as a 32-bit dll using Visual Studio 2015 (_MSC_VER 1900) Regrettably, I'm not set up for testing sqlite itself, and besides, the amalgamation is apparently too large for the VS debugger (it can never seem to break on the correct source line).