[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Behavior check: svnadmin load of revisions with no datestamp

From: Daniel Rall <dlr_at_collab.net>
Date: 2007-02-26 22:02:18 CET

On Mon, 26 Feb 2007, C. Michael Pilato wrote:

> It appears that today if you 'svnadmin load' a dumpfile that contains
> revision with no svn:date property, the resulting repository will, for those
> revisions, have that property (set to the time of the commit of the loaded
> revision).
> I can't decide if this is a feature or a bug, but I lean toward the latter.

IMO, this is definitely a bug, since it will result in a repository
with revision dates and numbers out of sequence, which impacts the
behavior of some operations which take a date or date range argument
(e.g. 'svn log -r {START_DATE}:{END_DATE}').

Subversion expects "svn:date" to be set -- it doesn't consider an
empty date to be valid data.

> If revisions lack a datestamp, it is almost certainly not accidentally so.
> So perhaps the load logic should be unsetting the svn:date property after
> committing revisions which did not carry that property in the dumpstream.

What effect does an empty date have on operations which accept a date
or date range?

Setting the date to the next millisecond after the date of the
previous revision seems better to me. While it has the side-effect of
creating data which was not specified in the dump file (which is
apparently consistent with current behavior), it keeps the repository
data in the format expected by Subversion's own operations.

  • application/pgp-signature attachment: stored
Received on Mon Feb 26 22:00:55 2007

This is an archived mail posted to the Subversion Dev mailing list.