On Tue, Sep 24, 2019 at 11:53 PM Alvaro Herrera <alvhe...@2ndquadrant.com> wrote: > > I think the danger is what happens if a version of your plugin that was > compiled with the older definition runs in a Postgres which has been > recompiled with the new code. This has happened to me with previous > unnoticed ABI breaks, and it has resulted in crashes in production > systems. It's not a nice situation to be in.
Indeed. > If the break is benign, i.e. "nothing happens", then it's possibly a > worthwhile change to consider. I suppose the only way to know is to > write patches for both sides and try it out. IIUC, if something like Heikki's patch is applied on older branch the problem will be magically fixed from the extension point of view so that should be safe (an extension would only need to detect the minor version to get a more useful error message for users), and all alternatives are too intrusive to be patckbatched.
