You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by "Miller, Eric" <Er...@amd.com> on 2007/11/01 16:14:43 UTC

RE: FW: Multi-path updates

> Ok - I'm now 100% sure.  Using the svn+ssh protocol the ra_session
takes
> 80-90% of the time in that loop.  Which makes sense since it is
opening
> and closing a full ssh session for every path specified.

I just filed issue #3003 for this.

I have a proof of concept written in perl that uses a new
svn_wc_crawl_revisions3() and am in the process of reworking
libsvn_client/update.c to utilize it as well.  I don't expect to need a
new update interface api - the change should be transparent.

On that note - would it be at all possible to get a development branch
so I can hack on this and other bugs?

Thanks,

Eric


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org


RE: FW: Multi-path updates

Posted by "Miller, Eric" <Er...@amd.com>.
> "Miller, Eric" <Er...@amd.com> writes:
> >   My edits have been against 1.4.2 - I have not taken a single look
at
> > the depth features of 1.5.  Does anyone know if --depth will
conflict
> > with this implementation?
> 
> Whoa -- you should *definitely* be working against the latest trunk
> code at all times.  Anything else will just cause you (and us) pain
> when you try to integrate your changes later.

I have no intention of submitting a patch against 1.4.2 code (unless you
want me to).  I will take the burden of integrating the changes against
the latest code base when I get to that point.  For now I have to get
this changes implemented against a known-good install base.

But back to my questions... =)


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org


Re: FW: Multi-path updates

Posted by Karl Fogel <kf...@red-bean.com>.
"Miller, Eric" <Er...@amd.com> writes:
>   My edits have been against 1.4.2 - I have not taken a single look at
> the depth features of 1.5.  Does anyone know if --depth will conflict
> with this implementation?

Whoa -- you should *definitely* be working against the latest trunk
code at all times.  Anything else will just cause you (and us) pain
when you try to integrate your changes later.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

RE: FW: Multi-path updates

Posted by "Miller, Eric" <Er...@amd.com>.
> I just filed issue #3003 for this.
> 
> I have a proof of concept written in perl that uses a new
> svn_wc_crawl_revisions3() and am in the process of reworking
> libsvn_client/update.c to utilize it as well.  I don't expect to need
a
> new update interface api - the change should be transparent.

I've made progress on this - I have a semi-working fix that I'm starting
to get ready to checkin to a devel branch.

Some background info:

First - I created svn_wc_crawl_revisions3 to take multiple paths,
starting and stopping the editor reporter on the first and last path
entries.

Next I updated the svn_client__update_internal routine to take a list of
paths and create the adm_access tree accordingly (update and update2
were modified accordingly).   The partial adm_access tree is passed to
the new crawler.  The result is a partially locked working copy tree and
a single ra_session and update editor reporter driven by only the paths
specified.

Currently I am simply using svn_path_get_longest_ancestor on the
user-supplied path list to determine the root for the adm_access set.  I
know this is deficient because path sets may not have a common root
unless translated into absolute paths first.  (Ex: svn up -N * */* does,
but svn up * ../* does not).

So my questions:

  What is the best way to find a common root path?  I was thinking of
converting everything to absolute paths first.

  My edits have been against 1.4.2 - I have not taken a single look at
the depth features of 1.5.  Does anyone know if --depth will conflict
with this implementation?

Thanks,

Eric M.


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org


Re: FW: Multi-path updates

Posted by Karl Fogel <kf...@red-bean.com>.
John Peacock <jo...@havurah-software.org> writes:
> Karl Fogel wrote:
>> (If you don't provide a password, a temporary one will be chosen for
>> you -- in either case, you can change your password yourself later via
>> <https://svn.collab.net/wpct/>.)
>
> Is the source for that available?  I don't see it in Google...

I dunno, I think maxb wrote it.  It sure should be available!

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: FW: Multi-path updates

Posted by John Peacock <jo...@havurah-software.org>.
Karl Fogel wrote:
> (If you don't provide a password, a temporary one will be chosen for
> you -- in either case, you can change your password yourself later via
> <https://svn.collab.net/wpct/>.)

Is the source for that available?  I don't see it in Google...

John

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: FW: Multi-path updates

Posted by Karl Fogel <kf...@red-bean.com>.
"Miller, Eric" <Er...@amd.com> writes:
> I just filed issue #3003 for this.
>
> I have a proof of concept written in perl that uses a new
> svn_wc_crawl_revisions3() and am in the process of reworking
> libsvn_client/update.c to utilize it as well.  I don't expect to need a
> new update interface api - the change should be transparent.

Thanks.

> On that note - would it be at all possible to get a development branch
> so I can hack on this and other bugs?

Sure, we've generally handed those out liberally.  Send a private mail
(quoting this one) to svnadmin{AT}svn.collab.net, giving your
tigris.org username and, optionally, an htpasswd-crypted password.
(If you don't provide a password, a temporary one will be chosen for
you -- in either case, you can change your password yourself later via
<https://svn.collab.net/wpct/>.)

Once you are notified that your account has been activated, just
create the branches you need in /branches (one branch per bug -- don't
mix different efforts into the same branch, please).  And I assume
you've read http://subversion.tigris.org/hacking.html already... :-)

Best,
-Karl

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org