> From: Sean Whitton <[email protected]> > Cc: [email protected], [email protected], > [email protected] > Date: Tue, 30 Jun 2026 14:14:13 +0100 > > Eli Zaretskii [29/Jun 7:38pm +03] wrote: > > If they use the released Emacs NN.1 instead, what's the difference? > > They will still see the regressions. > > One of us is missing something. We are talking about unreleased changes > on the emacs-NN branch. So, after the release of NN.2, say, the changes > on emacs-NN that will go into NN.3 but are not yet released. In that > situation, these changes will not get much testing because after the > release of NN.1 or at least after NN.2, most people will be either > running master or the released version.
That is true, but why is this crucial or even important for the issue at hand? Here's how this sub-thread developed: > > > > > > >> >> From: Stefan Monnier via "Emacs development discussions." > > > > > > >> >> <[email protected]> > > > > > > >> [...] > > > > > > >> >> - After NN.1 is released, I expect much fewer people run the > > > > > > >> >> `emacs-NN` > > > > > > >> >> branch, so any regression risks remaining undetected before > > > > > > >> >> the next > > > > > > >> >> (minor) release. If the regression is detected earlier, > > > > > > >> >> it'll likely > > > > > > >> >> be detected by someone running `master` and there's a > > > > > > >> >> chance that the > > > > > > >> >> fix will make it only to `master` if we don't notice the > > > > > > >> >> link to the > > > > > > >> >> previous bug-fix or don't notice that that previous bug-fix > > > > > > >> >> was > > > > > > >> >> installed in `emacs-NN`. > > > > > > >> > > > > > > > >> > The problems for which we install fixes on the release branch > > > > > > >> > after > > > > > > >> > NN.1 release come from 2 sources: > > > > > > >> > > > > > > >> Just to clarify: The paragraph I wrote above talks regressions > > > > > > >> due to > > > > > > >> new bug-fixes installed on the `emacs-NN` branch, not regressions > > > > > > >> present in NN.1. > > > > > > >> > > > > > > >> Rereading just now made me realize that it wasn't obvious at all. > > > > > > > > > > > > > > No, it wasn't. But anyway, bugs we fix on a release branch aren't > > > > > > > limited to regressions since NN-1.x releases, we also fix bugs in > > > > > > > earlier releases, as long as they are safe to fix. > > > > > > > > > > > > No, no, I'm really talking about regressions introduced by the > > > > > > bug-fix > > > > > > patches installed in `emacs-NN`. IOW about the problems that show > > > > > > up > > > > > > when what we seemed like a "safe bug-fix" isn't. > > > > > > > > > > Then I don't think I understand what you are saying. If those > > > > > regressions are reported while emacs-NN is still active, we will > > > > > first try to fix them there. If the problem such a regression causes > > > > > is not important enough, or if we already decided not to have any mode > > > > > NN.x releases, we will indeed fix it on master. > > > > > > > > > > So basically what you are saying is that we don't always pay enough > > > > > attention to decide that the fix should be on the release branch. If > > > > > that's your point, all we can do is be more vigilant. (I also don't > > > > > think I agree with you that this happens frequently enough to be a > > > > > problem, at least not lately. But I don't have any numbers to back > > > > > that up.) > > > > > > > > I'm saying they will tend to not be noticed because too few people run > > > > the code from that branch (after the NN.1 release), which means we have > > > > to be more conservative when trying to estimate whether a bug-fix is > > > > safe enough. > > > > > > I guess that depends on the workflows each one of us uses. I usually > > > try to reproduce on the release branch, as long as it is active (i.e., > > > as long as we haven't decided to have no more releases from it), even > > > if the bug is reported for the version on master, and if the problem > > > can be reproduced on the release branch, consider whether the fix is > > > safe enough to install there, rather than on master. YMMV, of course. > > > > I think the point is that people stop running emacs-NN as their > > primary/main/ordinary build of Emacs as soon as NN.1 is released. > > If they use the released Emacs NN.1 instead, what's the difference? > They will still see the regressions. Now, it's true that people running NN.k will not see regressions introduced on the emacs-NN branch after NN.k was released, but if those regressions are reported by people running master, and if we try reproducing them on emacs-NN before deciding whether the regression should be fixed there or on master, why does all of that matter? > >> Right now loads of people are running emacs-31 so regressions caused by > >> fixes we've installed there are much more likely to be detected. > > > > I would hope (and expect, TTTT) people to start using Emacs NN.1 as > > soon as it is released. That's what I do. My production sessions > > always run the latest official release or a pretest of the next > > release. > > I think you might be unusual in this regard. I think most people switch > to master once NN.1 is out, at least on the machine they do Emacs > development on. Then they will see on master the regressions introduced in emacs-NN, and we should remember to try the recipes on emacs-NN even if they are reported for master. Then we can decide on which branch to fix the regressions. What am I missing?
