[cvsnt] Lock Retry Interval and Retry Count

[Bo Berglund]

On Wed, 19 Jul 2006 07:52:38 -0700, "Michael Wojcik"
<Michael.Wojcik at microfocus.com> wrote:

>They agreed to upgrade the server a while back.  It just hasn't happened
>yet.  It's the usual story: they need a period when they have time to do
>it and the development teams can all cope with possible disruption if
>something goes wrong, and those two events rarely coincide.
>
>Tony's other suggestions (such as not committing the results of test
>builds) are good, but each development team here makes its own decisions
>about just what to put in CVS, and many have simply taken to throwing
>everything into the repository.  Not long ago I finally got rid of
>thousands of intermediate object files and the like from ours...

Those are really useless because eiter you have the build environemnt
and then these are recreated during build or you don't in which case
they are utterly useless and the only thing you need are the resulting
binaries....

We also keep the exe and dll binaries created in our CVS server, but
we do not commit these unless a binary has been sent out the door. And
in this case we *always* also set a tag on that particular snapshot
(including the binary).
The problem is RCS file growth for binaries, after a few hundred
commits of a fairly moderate exe file for example the CVS server has
to deal with RCS files of several hundred megabytes and this slows
down operations like tagging and log.
>
>On the other hand, when I started at my previous employer, there was no
>source change management at all - just lots of dated backups on file
>servers, floppies, and tapes.

Like we had around year 2000 when we realized we needed a version
control system and started shopping for one. We tested various
commercial packages but we soon wound up with CVSNT as a "test" system
since it was free and we were unused to version control systems. But
we have stayed with CVSNT ever since and it has served us well. :-)
Together with WinCvs it is really easy to get developers to use the
system.

>One of the first things I did there was
>port RCS to Windows and OS/2, and create a distributed RCS for OS/400.
>At least here at MF we're erring on the side of caution.


/Bo Berglund