Re: How to tell CVS that it should use the proper date/time

[Paul Sander]

>--- Forwarded mail from address@hidden

>Derek R. Price wrote:

>> Yes.  RCS_getdatebranch() in rcs.c stops when it finds the first date 
>> that is too old. From RCS_getdatebranch() in rcs.c: 


>Of course, this doesn't necessarily mean that that can't be fixed, but 
>any archive file that took advantage of it would no longer work quite 
>correctly with old versions of CVS and maybe RCS.

>I also suspect that there will be at least some administrators that 
>would rather we set at least the commit timestamps securely in 
>client/server mode.

In an RCS context, it makes sense not to store the file's timestamp.
The reason is because you can corrupt the ordering of the versions
so that fetching groups of files by timestamp produces configurations
that never existed before.

In a CVS context, this doesn't matter as much because the concurrency
algorithms keep the timestamps on the working files moving forward.  I
personally think that the time of the beginning of any given commit
should be recorded in every file that gets revved.  And the time should
be the time on the server, in case the clients are out of synch.  That
way you can't get a partial change if by checking out by timestamp you
happen to give a time that's in the middle of a commit.

>--- End of forwarded message from address@hidden