On Friday, April 7, Stephen Rasku wrote:
>
> >Don't do that. Working directories are not designed to be shared
> >between systems with incompatible file formats (e.g., DOS and Unix).
> >Put the repository on a shared file system if you must (and even then
> >it's better to use client/server CVS instead), but *NEVER* try to share
> >working directories.
>
> This seems like a bug to me
NOOOOOOO. This is not a bug. So what you are saying, is that CVS now
also needs to include the format, and deficiencies of all the files in
the CVS/ directory on the client side of things to be included in the
CVS protocol. The client manages the client CVS/ directories, and their
contents. The server manages the repo. Bloody hell, don't people
understand client-server operation and boundaries anymore?!?
You wouldn't expect Win2K to make use of a unix kernel now, would you?
Nor would you expect NT3.5 to make use of /etc/inetd.conf, or hosts.allow,
right? So why would you expect a unix program to make use of a DOS
programs *internal* state?!?
Besides, how are you managing all that cross mounting, multiple access to
the same place in the tree? How many developers are we talking about here?
Can't possibly be more than 5, if they are to get any decent work done...
> We want to do cross-platform development
> and we would like to compile the same source on all platforms without
> having to check it in first.
Why not check it in?
> We have the following structure for the
> each package in our system:
>
> pkg
> src
> tgt
> sol4 - For Sparc Solaris
> soli - For Intel Solaris
> win32 - For Windows
>
> This allows the same source to be compiled from the same location for
> multiple platforms.
So, you're saying that the "tgt" directory is within the CVS repo?
If so, then I'm telling you that you are making a mistake here. You're
going to run into trouble really quickly, unless there is a .cvsignore
file there to ignore that entry...
--Toby.
