one note: this patch can enforce a compatibility issues - a partially broken functions, where some badly written RAISE statements was executed newer.
I am not against this patch, but it should be in extra check probably ??
I'm not sure about what you mean by "it should be in extra check".
Or we have to documented it as potential compatibility issue.
Indeed, as a potential execution error is turned into a certain compilation error.
If this compatibility point is a blocker, the compilation error can be turned into a warning, but I would prefer to keep it an error: I'm quite sure I fell into that pit at least once or twice.
I prefer error, although there is possibility to control a force of exception - error or warning via extra plpgsql checks - this kind of error is strange - and stored procedures usually can be fixed later because plpgsql source is available.
There was a precedent with more precious query syntax check more times. Just it should be well documented.
