On Fri, Oct 10, 2008 at 11:18:51 -0400, David Roundy wrote: > Another possibility is to drop the human-friendly ChangeLog > generation.
Sure! For efficiency, we are already hard-coding the ChangeLog for released versions of darcs anyway (we still ought to do this for 2.1.0) > I don't know who actually uses it. Me :-) I don't think the tool needs to be part of darcs or our build system, but I find the underlying assumptions (that every patch needs to be explicitly marked) to be useful. So it would be nice if somebody could generalise the tool maybe, for example, by making it parse darcs changes output. I think I could use something like a darcs changes --xml --from-tag 2.0.1 > foo.xml generic_make_changelog foo.xml current_entries > I think the real problem here is the assumption that every darcs patch > must be explicitly handled in the changelog generation scripts, which > puts a huge burden on the maintainer (you). That's a relatively new > assumption, although I don't recall who introduced it. Mark requested it on http://bugs.darcs.net/issue705 and I implemented it. I find the assumption to be valuable, and I am willing to accept the associated costs. It would be nice if others were willing to share these costs, but in no way to do I want to impose them on other darcs hackers. What's the value in looking at each and every patch? It means we are less likely to forget to changelog something, which can happen without a disclipined approached. > It might make more sense to just let developers decide when some > change is worth mentioning in the changelog. Well, we always make the decision. The distinction is whether we make the decision explicit or not. Making the decision explicit is nice because (a) it makes it practical to enforce the discpline of looking at each and every patch and (b) it lets me know that the decision has already been made in the past, so that I don't have to make it again. The end result is more tedious (more busy work), but strangely easier (not having to think about it as much). > I see no reason why you should have to write 145 lines of code in > order to generate 16 lines of ChangeLog (which includes only 8 lines > of actual change text). Of course, some of those 145 lines (which is > the number of lines added) are just older lines that have gotten moved > around (since you also removed 103 lines, according to darcs), but > this is still a whole lot of churn for very little benefit. Yep. I could support a plain old text file being our ChangeLog. The benefit is that editing the ChangeLog becomes more intuitive for casual hackers. But I would still want the tool to be available on the side. -- Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow> PGP Key ID: 08AC04F9
signature.asc
Description: Digital signature
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
