[Mono-devel-list] Re: [Mono-patches] r36124 - trunk/mono/mono/mini

[Ben Maurer]

[am moving this to mono-devel-list, I should have sent it there in the
first place]

On Mon, 2004-11-15 at 15:36 +0100, Paolo Molaro wrote:
> On 11/15/04 Ben Maurer wrote:
> > Now that we are using subversion, it would be nice to take advantage of
> > atomic commits. This commit (r36124) depends on the commit you made 18
> > seconds earlier in metadata (r36123).
> 
> I know, I committed them that way on purpouse.


> BTW: atomic commits refer to the repository consistency, not to
> the consistency of the project under revision control.
Sorry, I should have used the word `changeset'.

> And make a mess of the commit message, including having svn log for a file
> include the changelog of other directories.
> I think having a clear history is more important than trying to make every 
> random svn revision buildable which is not possible anyway.

I had a talk with the guys in #svn about this, and how they changesets.
They take advantage of the feature. They pointed out the following
benefits they have seen from it:

      * It is easier to understand a single changeset (ie, a single
        email to mono-patches), as it has all the necessary context.
      * When looking at a log file, it is easier to find other changes
        that were necessary to make this set of changes. You don't have
        to go looking around for multiple changes around the same time.
      * It is easier to merge patches into other branches

The subversion guys seem to really enjoy the way they commit code. I
looked at one of their changesets via the online interface:

http://svn.collab.net/viewcvs/svn?rev=6443&view=rev

Note how the test and the change were committed at the same time. So if
I am asking `what is the history of the C code that was changed' I can
instantly find the test case.

> If the svn developers will add the ability to have different commit messages
> for different files in the same commit as other revision control systems do,
> we'll use them.

About svn log files becoming too big: I don't even think we are making
use of svn log much. When I talked on irc about it, the conclusion was
people make use of the changelog rather than svn log.

Also, if changes are truly related, having the information about the
related changes when you look for log information is valuable. Having
the logs be separate for each file seems to be separating valuable
information.

Now that we have switched to a system that has changesets, it seems
somewhat silly to not use them. The having bigger svn log files (which
are likely to hold relevant information!) seems a good trade for getting
changesets.

If Mono is going to follow a policy on changesets (either that you use
them or you don't -- I'd prefer using them and hope you will change your
mind, but will defer to policy) it would be nice to have this written in
the contributing section of the website.

-- 
Ben Maurer <bmaurer at ximian.com>