mirror of https://github.com/postgres/postgres
The SQL test added in this commit check a specific code path that had no coverage until now. When a TOAST datum is rewritten, toast_save_datum() has a dedicated path to make sure that a new value is allocated if it does not exist on the TOAST table yet. This test uses a trick with PLAIN and EXTERNAL storage, with a tuple large enough to be toasted and small enough to fit on a page. It is initially stored in plain more, and the rewrite forces the tuple to be stored externally. The key point is that there is no value allocated during the initial insert, and that there is one after the rewrite. A second pattern checked is the reuse of the same value across rewrites, using \gset. A set of patches under discussion is messing up with this area of the code, so this makes sure that such rewrite cases remain consistent across the board. Author: Nikhil Kumar Veldanda <veldanda.nikhilkumar17@gmail.com> Co-authored-by: Michael Paquier <michael@paquier.xyz> Discussion: https://postgr.es/m/CAFAfj_E+kw5P713S8_jZyVgQAGVFfzFiTUJPrgo-TTtJJoazQw@mail.gmail.compull/239/head
parent
60b64e6a31
commit
69f75d6714
Loading…
Reference in new issue