On 8/28/06, Antonio Petrelli <[EMAIL PROTECTED]> wrote:
Where is The Truth? :-P

In the implementation. :)

Something that could go wrong under CVS is that one file's commit
would fail, the others would succeed, and the repository would be in
an unexpected state. Smaller commits could be better, since it meant
there was less that could go wrong. SVN uses atomic committs, so
larger, multifile commits are not risky. It all succeeds, or it all
fails.

Under SVN, it's easy to roll things back to a particular commit, and
to reference the changes by the revision number. So if we are
implementing change XYZ, it's helpful to be able to say the change was
implemented in r1234567. And if we decde to roll back, we have an
implicit tag corresponding to the change.

Under CVS, I don't think I ever actually "rolled back" a commit. It
was too much work. Of I needed to undo something, I copy-and-pasted
the code forward. Under SVN, I often rolled back commits that didn't
work, by using SVN to copy forward the files that I want. (Not so much
here, but at work.)

Though, in either case, if the changes to the files are coupled, and
we aren't talking about a humungous number of CVS files, I'd commit
the changes together, for the same reason: "so it is clear what it is
changed in that file and why it has been done."

The evil practice is committing unrelated changes together, or mixing
stylistic and functional changes in the same commit.

-Ted.

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to