Ikumi Keita <[email protected]> writes: Hi Keita,
>> What I didn't understand in Keita's recipe is why one should >> uninstall and restart emacs before installing the new auctex version. >> Is that really needed because the presence of the old version causes >> the new version to be miscompiled? > > Maybe unintalling is overkilling. I'm worrying a bit what happens if > the user introuduced a package management system which forces loading > the AUCTeX package. (I don't have knowledge of package managers > including use-package, so I might be saying something silly.) I think the more elaborate package managers provide mechanisms for easily deferring package load until it is actually used in order to make emacs startup faster, so probably frequently auctex isn't loaded although the package is installed. Or are you talking about loading the package _after_ updating it? I think, package.el does that but I'm not sure. > Is it enough to start a new emacs session for updating with -Q option? Yes, I guess so, but I still don't understand what you think could go wrong when a user has an existing emacs session with auctex 13.3.x loaded and then updates to 14.0.0 in that session. I mean, except that auctex in this very session might be broken which is totally ok from my POV. Will an emacs restart not fix any issue? Your cautious receipe suggests that in the above scenario the newly installed auctex 14.0.0 will also be broken, even after restarting emacs. That can only happen with macro changes where byte-compilation with the wrong version of the macro would generate code that's incompatible with the new package version. And even then, a M-x package-recompile auctex RET followed by a restart should fix the issue. So if my thinking isn't wrong, I wouldn't add special documentation for the update from 13 -> 14. That updates, especially major version updates, may contain incompatible changes and may break the package in the current session should be considered normal. Bye, Tassilo
