Re: To ChangeLog or not to ChangeLog, that is the question

sb> It has been suggested that as mild an improvement in code sb> documentation as adopting full FSF Emacs-style ChangeLogs is sb> hurting XEmacs development. I don't think any patch should sb> patch should be applied to the code base without a ChangeLog sb> entry. I don't know how Richard manages to keep with the sb> minute details of every single patch, but I can't. I can sb> manage in the case of (relatively) simple patches. Large sb> patches are problematic. Given the debate about that, I'll just through my opinion on this, based on our own experience here with the software we develop. What we usually do is the following: - Only do global commits. Local commits are evil, because there is very little way of tracking them, and/or forcing people to add a Changelog Entry. - commits are usually made only once in a while, when people need to syncronize things. - ChangeLog entries are *required* for each commit. They usually consist in the name of the developper who commits, date, list of modified files, and main reasons why they were modified. - Details about why a file was modified are in the log for that particular file. I usually do a global cvs diff before doing a commit. That helps me creating almost automatically the changelog entry, and find out what I'm supposed to put in the local log file. In your case, though, some patches come from people who don't commit things themselves. So I think the scheme could be extended so that the main changelog entry would contain the list of files modified by each patch, and the commit log entry the list of patches (i.e. originator etc...) for a given file. That sure can be a real pain, but it has helped us a lot in the past tracking who was reposible for what here (the project I work on is more than half a million lines of C++ code, with about 15 developpers). As to vote on whether there should be or not a ChangeLog, it basically depends on whether they are of any use to the maintainers. If there bringing pain for little benefit, I'd say drop them. On the other hand, they have prooved very usefull here, provided that clear rules for changelog/local logs entries were properly followed by all developpers involved... Richard. -- |-------------------------------------------------------------| | Richard Cognot | Proceed, with fingers crossed... | | | /\^^/\ | | <cognot(a)ensg.u-nancy.fr&gt; | `' | |-------------------------------------------------------------| | http://www.ensg.u-nancy.fr/~cognot | |-------------------------------------------------------------|