On 04.06.21 06:28, Julien Rouhaud wrote:
Yes, but we have a lot a examples of functions without pg_nodiscard and callers still explicitly ignoring the results, like fsm_vacuum_page() in the same file. It would be more consistent and make the code slightly more self explanatory.
I'm not clear how you'd make a guideline out of this, other than, "it's also done elsewhere".
In this case I'd actually split the function in two, one that returns void and one that always returns a value to be consumed. This overloading is a bit confusing.
