cvup mirrors and exact CVS dates

[Cristian KLEIN]

Erik Cederstrand wrote:
> Hi!
> 
> I'm chugging along with the performance tracker project[1]. I have a
> script constantly building jails using the newest CURRENT sources, and
> since reproduceability is one of the goals, I have some questions
> regarding cvsup mirrors and CVS in general (hope it's OK on this list):
> 
> 1) To identify the jails, I name them after the CVS date I put in the
> cvsup file. Now, if the date is now() (converted to UTC), can I be sure
> that any given mirror (retrieved by fastest_cvsup) has the same set of
> files? Or will some mirrors be behind and happily give me sources that
> are older than what the main CVS server would report for that specific
> CVS date?

Suppose the mirrors we are talking about are up'n'running, some mirrors may be
as much as 1 hour behind with the source tree.

> 2) On the same note, is there any way to uniquely identify a global
> revision of the FreeBSD source tree, comparable to revisions in
> Subversion? From what I understand, there is no way to know if e.g. CVS
> dates 2007.10.25.13.45.00 and 2007.10.25.13.46.00 represent the same
> source code, except updating to both dates and seeing if any files change.

This is my understanding too. Each file has a separate version number and there
is no global revision number. If you still require such a feature, how about
taking max(mtime()) as a global revision number?