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

Re: svn commit: r1547866 - /subversion/trunk/subversion/libsvn_fs_fs/rep-cache.c

From: Stefan Sperling <stsp_at_elego.de>
Date: Thu, 5 Dec 2013 14:04:08 +0100

On Wed, Dec 04, 2013 at 08:09:32PM +0000, Philip Martin wrote:
> Philip Martin <philip.martin_at_wandisco.com> writes:
>
> > Stefan Sperling <stsp_at_elego.de> writes:
> >
> >> Oops, I pasted an old version of my diff. Here's the final one,
> >> mostly with adjusted comments.
> >>
> >> Overall, I like your trick with creating an empty file better.
> >>
> >> What do you think about making 'svnadmin create' create rep-cache.db?
> >
> > My new code makes rep-cache.db have umask permissions but I see now that
> > is probably not be what we want. When commit creates new revision and
> > revprop files it propogates the permissions from the r0 revision file.
> > I think that when we create the rep-cache we should do the same.
>
> It's rHEAD rather than r0 but the aim is to make all revision files have
> the same permissions so there should be no difference which is copied.
> However revision files are not writeable so simple copying won't work
> for the rep-cache.

I think we do want umask permissions. The rep-cache is written by
the server process, and there are cases where the group write bit
must be set for this to work (e.g. in some svn+ssh setups). In any
case the server process will have the proper umask configured, so
we can just use it.

Revision files are usually read-only so the story is different there.
Received on 2013-12-05 14:05:08 CET

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