Last month, I wrote about Charles Choi’s post on the the on package-install-upgrade-built-in variable and how it controls the upgrading of builtin packages. The TL;DR is that when it’s set to t, package.el treats builtin packages as if they had been installed like any other 3rd party package for upgrading purposes.
My first inclination was to wonder why anyone would not want to have it set to t. I could imagine some outlying cases where it didn’t make sense but mostly I’d have expected that upgrading builtin packages was what most users wanted.
Apparently not. Over at the Emacs subreddit, jonas37 asks if people are setting package-install-upgrade-built-in and the comments are revealing. It’s hard to imagine the answer would be controversial but lots of people have strong feelings on the matter.
Take a look at the comments to see what I mean. The answer, as it usually is, is that “it depends”. In short, it’s hard to automate the decision. Sometimes you want to upgrade the builtins and sometimes you don’t.
It’s easy to take the position that your editor, whatever it is, should take care of things for you, but people who do take that position are the first to complain when the algorithmic decision is not to their liking. Sadly, sometimes we just need to engage with Emacs and tell it what we want it to do about upgrading our packages. For many cases. setting apackage-install-upgrade-built-in to t is probably the right decision but sometimes not. It, as they say, depends.