k> I think he was talking about the client attempting to auto-detect
k> such things for the case when the user neglected to use the `svn
k> foo' commands. It would be great, as long as the client confirms
k> its deductions with the user.
Yep. This would be a very valuable thing. CVS gets this half-right
with the cvsignore file. To complete it, what you want is the
*opposite* of a cvsignore file - a set of patterns that, if any of
them match and aren't in the repository, cause an operation to fail.
Why is this useful? Well, let's say you check out a view of the
repository and spend a month banging on it, then want to check your
changes in. Along the way, you've created a few source files that
you've forgotten to "svn add". If you do a checkin with those files
missing, the image in the repository probably can't be built by anyone
else.
This is more of a problem with large projects or in ones where
individuals make a significant number of changes than in noddy little
projects; it happens fairly often at my workplace, where we use P4.
We've hacked together a script to find "missing" files, but people
have to remember to run this by hand. It would be much better for our
collective sanity if the client did this check for us automatically,
since it has to scan directories anyway.
Now, this would be trivial to add to an extensible client using some
Guile code or whatever, so it's not at all clear to me that it needs
to go into the base client. I assume that the client will be
extensible at some point? And that a library of "standard extensions"
will be packaged with it?
<b
--
Let us pray:
What a Great System.
Please Do Not Crash.
^G^IP@P6
Received on Sat Oct 21 14:36:05 2006