Date: Sun, 16 Jan 2000 10:53:31 +0530 From: Greg Lehey <grog@lemis.com> To: "Matthew N. Dodd" <winter@jurai.net> Cc: "Matthew N. Dodd" <mdodd@FreeBSD.ORG>, cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: How often to commit? (was: cvs commit: src/sys/i386/isa if_ex.c) Message-ID: <20000116105331.L1938@mojave.worldwide.lemis.com> In-Reply-To: <Pine.BSF.4.21.0001151352220.462-100000@sasami.jurai.net>; from winter@jurai.net on Sat, Jan 15, 2000 at 01:55:52PM -0500 References: <20000115130818.G349@mojave.worldwide.lemis.com> <Pine.BSF.4.21.0001151352220.462-100000@sasami.jurai.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Saturday, 15 January 2000 at 13:55:52 -0500, Matthew N. Dodd wrote:
> On Sat, 15 Jan 2000, Greg Lehey wrote:
>> When I see messages like this, all containing relatively small
>> modifications, I wonder if we should come to some agreement on how
>> often a commit should be made for modules on which we're actively
>> working. I'm not picking on mdodd (well, only in the sense that I
>> noticed these three commits); I'm just wondering. I tend to go to the
>> other extreme myself, testing changes in some detail before making a
>> mega-commit. I suspect that the truth lies somewhere between these
>> two extremes. Would anybody like to discuss where it might be?
>
> In the first commit I fixed the driver so it would work. I broke it back
> when I blindly converted it to newbus; my tester went away and I didn't
> get any real feedback. Since I now have a card I can fix and test myself.
>
> In the second commit I implemented new functionality.
>
> In the third I did some commetic cleanups.
>
> I've been fussed at for doing one big commit before and I've been burned
> before by doing massive changes (if_ep).
>
> So, if you're just using me as an example to start a conversation thats
> fine, but be aware that this specific case is a really bad example.
As I say, I chose this example simply because it was three commits in
10 minutes, no other reason. I'm not criticizing it, I'm just using
it as a basis for discussion. Based on these criteria, it seems quite
a reasonable example. My megacommits include all three of these
components. If you had committed all three fixes at once, it wouldn't
have made much difference to the end result.
A couple of further thoughts:
1. Warner mentions that it keeps people's attention. I'll agree with
that. I've been following his commits with great interest, but on
the other hand I've also been left with the feeling "why don't you
tell me when it's all over?".
2. Currently I'm travelling, and sometimes I go a couple of days
before I can pick up my mail. When I do, I find up to 500 commit
messages through which I have to struggle. I've already made a
few suggestions about how to structure the messages so we can pick
out the important ones and ignore the "fix typo" kind of message.
Fewer commits would also help.
3. Part of the idea of committing to the tree is to get people to
help in the testing. This would work if I were to commit Vinum
fixes more often than once a month. On the other hand, I don't
think that many people got much testing done between your three
commits :-)
In general, the replies I've seen so far tend to confirm my opinion
that the middle ground is better than either extreme, but I'm open to
further discussion.
Greg
--
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000116105331.L1938>