[cvsnt] Re: "cvs update" avoiding files with local changes?

[Johan Holmberg]

Gerhard Fiedler <lists at connectionbrazil.com> writes:

> Johan Holmberg wrote:
> 
> > I should perhaps have explained my situation in more detail. My
> > scenario (in this case) is to use "cvs update" in a "batch setting" to
> > update all files in a whole directory tree. The files are *self
> > contained*, so the usual dependency problems don't arise. When there
> > is un-committed local changes, it is for a good reason. The most
> > important thing (in batch mode) is to *avoid* overwriting these
> > changes, not even with a successful merge. I want those changes to be
> > managed interactively later.
[...]
> > I have thought about switching to some "mirroring software". It would
> > probably fit nicely with the "batch mode". But then I would loose the
> > nice things with CVS when working in "interactive mode": using diff,
> > accessing file history, commiting changes, etc.
> 
> I don't quite understand this, especially how a mirroring software would
> come into play. As I wrote in my previous message, in situations that may
> be similar to yours (I don't really understand your work flow yet),
[...]

Concerning the workflow: I develop a "test engine" used by 20 compiler
developers. The system has around 10_000 testcases. Normally the
developers get the files via the test engine without even having to
know that the testcases are stored in CVS (what I called "batch mode"
earlier).

Occasionally a developer is forced to make local changes to a
testcase. These changes are used locally for a while, and later
considered for addition to the central CVS archive.

In the meantime it is important that the developer can get other
changes from the server (like "cvs update" would do). But the locally
changed files should be left alone (if they happens to be changed on
the server too).

So the usage is 98% "batch mode" (the developer doesn't care about
CVS). But in the other 2% of cases CVS comes in very handy.  I have a
homegrown mirroring solution that can solve the "batch mode" for me,
but if we used it, we would be forced to handle the 2% in a different
(and worse) way.

Returning to my original question: my current understanding and
conclusion is that no variant of "update" exists, that avoid updating
conflicting files.

Maybe I should hack the CVS source :-)


/Johan Holmberg