Re: CVS checkouts of *.in and *.am files

[Derek R. Price]

Harlan Stenn wrote:

> Folks have recently discussed things like "egcs_update" and other
> related scripts that attempt to check out/update certain "master" files
> before doing a checkout of "derived" files.

> Before saying something stupid on the CVS list, I thought I'd try my
> idea out here!
>
> A possible solution to this problem might be:
>
> - on checkin, check the files in with the timestamp on the "local" file
>
> - on checkout/update, get the list of files that will be "messed with"
>   and simply do the work in oldest -> newest order
>
> The questions that come to mind are:
>
> - Will this work?

Yes, as long as the client and server can agree on the time.  Since they
probably won't, for most pratical purposes, the commit may have to go like
the co you proposed, and check in from oldest to newest.  This part of the
algorithm may still need some work.


> - Will it break anything?

Except as above, no.


> - Would it be difficult to do?

Most likely.  At least, it'd still take time for me to do right.  CVS is
designed to operate a directory at a time and this has become integral to
the client/server communication as well.  It's possible the update step
could be broken in two, where the first step constructs the list of files
and the second does the actual work, but this probably has some
implications worth a bit more consideration.

Derek

--
Derek Price                      CVS Solutions Architect ( http://CVSHome.org )
mailto:address@hidden     OpenAvenue ( http://OpenAvenue.com )
--
I will not charge admission to the bathroom.
I will not charge admission to the bathroom.
I will not charge admission to the bathroom...

          - Bart Simpson on chalkboard, _The Simpsons_