Mitch Davis schrieb:
> [Apologies if this is a repost, posting thru egroups.com has failed twice]
>
> Hello,
>
> I'm a long-time satisfied user of CVS remotely via
> SSH. I would like to enhance CVS to overcome a
> particular deficiency, but first I wanted to hear
> what the CVS "pros" think. That means you! :-)
>
> The basic problem is that detecting file changes
> by storing timestamps in the CVS/Entries file is
> prone to error.
>
> When the CVS client "update"s a file, the
> timestamp of that file is stored in CVS/Entries.
> When a new update is performed (or something else
> like the up-to-date check when checking in), the
> CVS client compares the stored timestamp with the
> file timestamp, and uses this as an indication of
> whether it should contact the server and request
> the unchanged version.
>
> The problem is that just using the timestamp is a
> poor indicator of whether the file has changed.
>
> - Due to copying between machines (Windows!), the
> timestamp may not be the same any more, although
> the data is.
>
> - The file may have changed (somehow) but the
> timestamp is the same (much less common).
>
> My suggestion is, instead of storing the timestamp
> in CVS/Entries, the CVS client stores the MD5 of
> the file. And the up-to-date check performs an
> MD5 of the file. If they don't match, the CVS
> client knows beyond all doubt that the file has
> changed, and can contact the CVS server.
>
> To see why this is needed, imagine you have a 50mb
> repository, you're connected over a modem, and all
> your files have been touched. With the present
> timestamp system, "cvs up" is going to take a
> LOOOONG time... :-)
>
> Making this change seems worthwhile, doesn't seem
> overly hard, doesn't change the CVS paradigm and
> doesn't change the CVS client/server protocol.
> (And CVS already includes the MD5 code...)
>
> So could people tell me whether or not
> this is a good idea please?
>
> All responses will be most gratefully received.
>
> Regards,
>
> Mitch.
> --
> | mailto:[EMAIL PROTECTED] | Not the official view of: |
> | mailto:[EMAIL PROTECTED] | Australian Calculator Opn |
> | Certified Linux Evangelist! | Hewlett Packard Australia |
As I already wrote in a previous posting I think it is really a good idea.
Right now I have seen a problem in it: CVS Windows clients automatically
convert text file line endings from LF to CRLF and vice versa. So checksums
computed on a Windows client will bedifferent from checksums computed on the
(*nix) server. (I don't know how the WinNT CVS server stores textfiles.)
Best regards
Martin
