On Mon, Aug 01, 2011 at 10:16:53PM +0400, Konstantin Kolinko wrote:
> 2011/8/1 Stefan Sperling <stsp_at_elego.de>:
> > On Mon, Aug 01, 2011 at 04:38:46PM +0200, Markus Schaber wrote:
> >>
> >> Hi,
> >>
> >> Von: Stefan Sperling [mailto:stsp_at_elego.de]
> >> > On Mon, Aug 01, 2011 at 04:12:41PM +0200, Markus Schaber wrote:
> >> > > Can subversion cope with repositories and working-copies on
> >> read-only
> >> > media?
> >> > >
> >> > > Maybe some Live-Demo CD bringing an example SVN repository?
> >> >
> >> > No.
> >>
> >> So subversion needs write access to both repository and working copies,
> >> even if the user performs only operations which he "thinks" to be
> >> read-only like "status", "diff", "compare to revision" for working
> >> copies, and "repo browsing" and "compare to revision" for repositories?
> >
> > Yes. For all read opreations it needs to create read-locks to prevent other
> > clients from writing at the same time. Another example is that during diff
> > it needs to create temporary files to write diffs to.
> >
> > Repo browsing (e.g. svn ls URL) will probably work as long the repository
> > is on a writable filesystem.
> >
>
> It is written in SVN book that repository that uses FSFS should work
> on read-only media.
> Is it something that was broken recently?
>
> - see "Usable from a read-only mount" row in the table in the middle of
>
> http://svnbook.red-bean.com/nightly/en/svn.reposadmin.planning.html
Sorry, I was mistaken there.
On IRC Bert Huijben and Daniel Shahaf corrected me:
<Bert> stsp: (re: users@) read only operations on a fsfs repository
don't obtain a lock (and don't need write access). Only commit and
revprop changes should need write access. (I don't know what is required
for bdb). We don't use read locks, except for the sqlite (and maybe bdb)
builtin ones.
<danielsh> stsp: and packing
Received on 2011-08-02 01:29:38 CEST