You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by "D.J. Heap" <dj...@shadyvale.net> on 2003/07/16 02:51:32 UTC

libsvn_fs rewrite

What all is being addressed in the libsvn_fs rewrite that cmpilato (I 
think?) is working on?

Some of the issues that look like they could be included in the 
libsvn_fs rewrite, if I understand things correctly, are:

Reduce transaction use in BDB.
Atomic renames.
Some work to help increase changeset support.
Some work to help support different SQL backends.
Others?

It seems unbelievable (although it would be wonderful) that all of these 
are being addressed by the rewrite...it *is* a rewrite of libsvn_fs, 
isn't it?  I'm just curious what's coming and what issues it is going to 
free up to be worked on since I've seen it mentioned a few times but 
haven't really seen any details.

DJ



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

Re: libsvn_fs rewrite

Posted by cm...@collab.net.
Branko Čibej <br...@xbc.nu> writes:

> >I originally wasn't planning to tackle this, but yesterday I got so
> >frustrated with our transaction (ab)use in the BDB that I started
> >plowing through Berkeley docs to see what I needed to do to remedy the
> >situation.  So... "maybe".
> >
> Good luck. I tackled this a couple of times, then decided to wait for
> gat's FS refactoring before even thinking about it again. Please holler
> if you find a silver bullet I overlooked.

You know, today I came to basically the same sort of despair.  So
instead of pushing forward on this front, I'm choosing a different
battle.  Glenn Thompson (and an SQL backend), here I come (as time
allows).

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


Re: libsvn_fs rewrite

Posted by Branko Čibej <br...@xbc.nu>.
cmpilato@collab.net wrote:

>"D.J. Heap" <dj...@shadyvale.net> writes:
>
>  
>
>>Reduce transaction use in BDB.
>>    
>>
>
>I originally wasn't planning to tackle this, but yesterday I got so
>frustrated with our transaction (ab)use in the BDB that I started
>plowing through Berkeley docs to see what I needed to do to remedy the
>situation.  So... "maybe".
>
Good luck. I tackled this a couple of times, then decided to wait for
gat's FS refactoring before even thinking about it again. Please holler
if you find a silver bullet I overlooked.


-- 
Brane Čibej   <br...@xbc.nu>   http://www.xbc.nu/brane/


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

Re: libsvn_fs rewrite

Posted by cm...@collab.net.
"D.J. Heap" <dj...@shadyvale.net> writes:

> Reduce transaction use in BDB.

I originally wasn't planning to tackle this, but yesterday I got so
frustrated with our transaction (ab)use in the BDB that I started
plowing through Berkeley docs to see what I needed to do to remedy the
situation.  So... "maybe".

> Atomic renames.

Not likely.  Depends on how good I feel about the rest of the changes.
And if I can get it cheaply.  Philip Martin and Bitt Tutt had code for
this support, so I might be able to just to work (read: "crowbar")
their changes into my new code.

> Some work to help increase changeset support.
> Some work to help support different SQL backends.

Nuh-uh.  I mean, yes, something I do might help changeset support down
the road, but I have no explicit goal there.  

The main problems I'm tackling are:

   - issue #1248 - fixes to the copy ID inheritance algorithm
   - issue #1003 - fixes to svn_fs_revisions_changed()
   - fixing some inefficiencies in specific uses BDB transactions
   - (possibly) fixing our busted BDB transaction usage policy
   - (possibly) implementing successor pointers in the DAG nodes
   - (possibly) solving the WebDAV caching problems
   
In the meantime, I can be heard groaning about how our lazy copy
model might wind up being more trouble than its worth...

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

Re: libsvn_fs rewrite

Posted by "D.J. Heap" <dj...@shadyvale.net>.
Sander Striker wrote:

[snip]

> 
> No.  It's straightening out some issues.  Not a full rewrite.  We couldn't
> do that to the poor guy.  Nor to ourselves, since that would put us
> back quite a bit, since no doubt new bugs would be introduced etc, etc.
> 
> 
> Sander
> 

Ah, I see, that makes sense, thanks.  There are some schema changes 
coming, though, right?  What are the problems being addressed?

DJ



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

RE: libsvn_fs rewrite

Posted by Sander Striker <st...@apache.org>.
> From: D.J. Heap [mailto:dj@shadyvale.net]
> Sent: Wednesday, July 16, 2003 4:52 AM

> What all is being addressed in the libsvn_fs rewrite that cmpilato (I 
> think?) is working on?

C Mike is, yes.
 
> Some of the issues that look like they could be included in the 
> libsvn_fs rewrite, if I understand things correctly, are:
> 
> Reduce transaction use in BDB.

Not in the branch.

> Atomic renames.

CMike?  Did it go in or not?

> Some work to help increase changeset support.

??

> Some work to help support different SQL backends.

This is something Glenn Thompson is working on.

> Others?
> 
> It seems unbelievable (although it would be wonderful) that all of these 
> are being addressed by the rewrite...it *is* a rewrite of libsvn_fs, 
> isn't it?

No.  It's straightening out some issues.  Not a full rewrite.  We couldn't
do that to the poor guy.  Nor to ourselves, since that would put us
back quite a bit, since no doubt new bugs would be introduced etc, etc.


Sander

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

RE: libsvn_fs rewrite

Posted by Tim Hawkins <ti...@timhawkins.co.uk>.
I would like to become involved in this process, as I am very interested in
this component for use in a standalone mode. 

-----Original Message-----
From: D.J. Heap [mailto:dj@shadyvale.net] 
Sent: 16 July 2003 03:52
To: dev@subversion.tigris.org
Subject: libsvn_fs rewrite

What all is being addressed in the libsvn_fs rewrite that cmpilato (I 
think?) is working on?

Some of the issues that look like they could be included in the 
libsvn_fs rewrite, if I understand things correctly, are:

Reduce transaction use in BDB.
Atomic renames.
Some work to help increase changeset support.
Some work to help support different SQL backends.
Others?

It seems unbelievable (although it would be wonderful) that all of these 
are being addressed by the rewrite...it *is* a rewrite of libsvn_fs, 
isn't it?  I'm just curious what's coming and what issues it is going to 
free up to be worked on since I've seen it mentioned a few times but 
haven't really seen any details.

DJ



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



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