Greg Stein <gstein@lyra.org> writes:
> > AFAIK, It's nothing but a convenience at this point (and my main reason
> > for pulling/reworking it would be that it's not as accountable as the
> > --revprop change method (yes, it writes changes to a file, but it
> > doesn't send a commit message).
>
> Agreed. I remember harping on Karl for similarly editing log messages
> "behind the scenes". His "justification" for not using --revprop was that
> the commit message didn't send a diff.
>
> But what it *does* send is more than the CGI.
>
> Oh, note that the CGI can do its magic even while revprop changes are
> disabled (no hook in place). But for the SVN repos, we *have* that hook,
> so I think the CGI ought to go.
-0. The CGI is a simpler interface that doesn't even require a
Subversion client (quite nice if you're, say, on your wife's machine
and want to see what's changing in the repository). You don't even
have to remember the revision number if the commit is within the last
20 or so (whatever the script displays).
If the script goes, then we should toss 'svnadmin setlog' for the
same reasons.
I'd rather see the script have an option to use either 'svnadmin
setlog' or 'svn pset --revprop' -- that way we can still have the ease
of the interface, but also get the commit mails.
Alternatively, just make 'svnadmin setlog' hit the filesystem via the
svn_repos() interfaces so it too can hit the hooks.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Oct 6 21:26:37 2003