On 9/19/07, David Glasser <glasser@davidglasser.net> wrote:
> On 9/19/07, Malcolm Rowe <malcolm-svn-dev@farside.org.uk> wrote:
> > On Wed, Sep 19, 2007 at 11:26:20AM -0700, David Glasser wrote:
> > > Three property names are defined in libsvn_fs/fs-loader.h starting
> > > with SVN_FS_PROP_TXN_: SVN_FS_PROP_TXN_CHECK_LOCKS,
> > > SVN_FS_PROP_TXN_CHECK_OOD, and SVN_FS_PROP_TXN_CONTAINS_MERGEINFO.
> > >
> > > Shouldn't they be SVN_FS__PROP_TXN_*?
> > >
> >
> > No, I think the first two are part of a public API, though I haven't
> > time to check just now.
>
> But they're defined in libsvn_fs/fs-loader.h... There's
> SVN_FS_TXN_CHECK_OOD and SVN_FS_TXN_CHECK_LOCKS in svn_fs.h but that
> is different (bits instead of prop names).
Made this change in r27022.
> > > (Also, I don't think contains-mergeinfo is necessary: it's only used
> >
> > (agreed).
> >
> > > by FSFS, and you should be able to just see if the mergeinfo file
> > > exists on disk instead, no?)
> > >
> >
> > Sure. I'd actually like to move the mergeinfo information -- currently
> > stored in a 'mergeinfo transaction property' file -- into the
> > already-existing transaction properties hash (since I see no reason why
> > we need a separate file). So we'd need to read that instead, but sure.
>
> I guess arguably there's some efficiency argument being made that you
> only have to read and dump the mergeinfo each time a path is added
> during the transaction build phase, instead of the whole txnprops hash
> (log message, etc). Of course the hash build is already quadratic...
Still considering this.
--dave
--
David Glasser | glasser_at_davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Oct 8 19:22:44 2007