Discontinuing the use of SVN.

[Robert Murphy <rm...@numediaprod.com>]

After much frustration (and yelling) I have been told to remove SVN from our
production pipeline after the latest round of ‘Checksum mismatch’ errors. We
will have to now migrate back to our older commercial RCS system. This
problem keeps cropping up and nothing that I have done so far appears to
help. 

 

Thanks for your earlier assistance,

 

Rob.


-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.322 / Virus Database: 267.5.2 - Release Date: 6/3/2005
 

Re: Discontinuing the use of SVN.

[Robert Sfeir <yo...@gmail.com>]

But I've also gotten them when a file is out of sync with the
repository somehow.  This used to happen, but no longer does, when
using a plugin to an IDE which used the javaSVN library.  This has
since been fixed.  It is entirely possible that something like
tortoise or other clients can cause this problem.

R

On 6/3/05, Ben Collins-Sussman <su...@collab.net> wrote:
> 
> On Jun 3, 2005, at 4:34 PM, Robert Murphy wrote:
> 
> > After much frustration (and yelling) I have been told to remove SVN
> > from our production pipeline after the latest round of 'Checksum
> > mismatch' errors.
> 
> Sorry about that.  The checksums exist to point out hardware problems
> with network, disk, and memory.  I hope your hardware is okay!
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: users-help@subversion.tigris.org
> 
>

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

RE: Discontinuing the use of SVN.

[Dale Worley <dw...@pingtel.com>]

> From: Paul Wayper [mailto:paul.wayper@anu.edu.au]
>
> This is probably a newbie question, but why not have two
> working copies,
> one on each machine, checked out separately?

No doubt that's better.  I started working the way I did because I was
already syncing two complete copies of my home directory anyway, so I
figured I could keep the Subversion WCs synced in the same way.  And
generally, it worked OK.

I was just attempting to emphasize that the *only* fragility I've ever seen
in Subversion is that I could damage my WCs when doing this, but that the
damage could have been fixed fairly easily. When someone else posted that
his organization has seen the same problem, it suggests that eliminating
this weakness of Subversion might substantially improve its robustness (in
the real world, where people abuse their WCs).

Dale


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

Re: Discontinuing the use of SVN.

[Paul Wayper <pa...@anu.edu.au>]

Dale Worley wrote:

>I used to get these messages regularly.  I never really tracked them down,
>but eventually it became clear that the central problem was that I was
>maintining two copies of the WC on two machines and using Unison to keep
>them in sync.  In theory, this isn't a problem if you maintain discipline,
>only use one WC at a time, and do a re-sync whenever you change your
>attention from one WC to the other.  But in practice, you forget to do that,
>and Unison can only partially propagate the changes.
>  
>
This is probably a newbie question, but why not have two working copies, 
one on each machine, checked out separately?  Then just remember to 
commit on the machine you're leaving and update on the machine your 
arriving at.  This sounds like exactly the sort of thing that Subversion 
is supposed to handle perfectly.  Is there a good reason for trying to 
have two copies of the same working copy?  If you can run Unison, you 
can probably run an ssh user@remotehost "cd work/projectx ; svn commit"...

Or am I missing something?

Have fun,

Paul

-- 
-- Paul Wayper at ANU - +61 2 6125 0643


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

Re: Discontinuing the use of SVN.

[Martin Probst <ma...@x-hive.com>]

Hi,

> That's true, Subversion could in theory recover from this.  However,
> our feeling has been that the .svn area is sacred: if you can't trust
> the text-base/ files, then you can't trust the entries file either --
> all bets are off.

I remember I used to wedge my .svn directories using TortoiseSVN - it 
was a bug related to the fact that Tortoise cannot change the path or 
something - I don't know if it's fixed by now.

However, this happened every now and then, and in the endeffect the only 
solution (without manually editing entries fiels etc.) was to delete the 
whole SVN checkout and co again.

Instead of dropping the whole repository, some forced fixup of the .svn 
directories might be nice, IMHO.

Regards,
Martin

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

RE: Discontinuing the use of SVN.

[Dale Worley <dw...@pingtel.com>]

> From: kfogel@newton.ch.collab.net
>
> That's true, Subversion could in theory recover from this.  However,
> our feeling has been that the .svn area is sacred: if you can't trust
> the text-base/ files, then you can't trust the entries file either --
> all bets are off.

I agree in principle.  But as I said, I ran into a remarkable number of
troubles that revolved around text-base being out of sync with entries.  And
though you can't *trust* entries, pretty much by definition, you can't make
the situation worse if you update text-base to match what entries says it
should contain.

And as you see, other users are running into this, too.

Dale


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

Re: Discontinuing the use of SVN.

[kf...@collab.net]

"Dale Worley" <dw...@pingtel.com> writes:
> As far as I know, all instances of "Checksum mismatch" were when my
> text-base file contained a different version than the entries file said it
> was.  There were also some other times when empty-file got deleted.  (I
> don't know why.)
> 
> What was particularly annoying about these situations was that they were, in
> theory, quite mendable -- a program could easily have detected the incorrect
> text-base files and queried the repository for the correct versions.  (This
> is a consequence of the fact that the text-base is just a cache.)  Sometimes
> I did this manually, and it worked as expected.

That's true, Subversion could in theory recover from this.  However,
our feeling has been that the .svn area is sacred: if you can't trust
the text-base/ files, then you can't trust the entries file either --
all bets are off.

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

RE: Discontinuing the use of SVN.

[Dale Worley <dw...@pingtel.com>]

> From: Ben Collins-Sussman [mailto:sussman@collab.net]
>
> > After much frustration (and yelling) I have been told to remove SVN
> > from our production pipeline after the latest round of ‘Checksum
> > mismatch’ errors.
>
> Sorry about that.  The checksums exist to point out hardware
> problems
> with network, disk, and memory.  I hope your hardware is okay!

I used to get these messages regularly.  I never really tracked them down,
but eventually it became clear that the central problem was that I was
maintining two copies of the WC on two machines and using Unison to keep
them in sync.  In theory, this isn't a problem if you maintain discipline,
only use one WC at a time, and do a re-sync whenever you change your
attention from one WC to the other.  But in practice, you forget to do that,
and Unison can only partially propagate the changes.

As far as I know, all instances of "Checksum mismatch" were when my
text-base file contained a different version than the entries file said it
was.  There were also some other times when empty-file got deleted.  (I
don't know why.)

What was particularly annoying about these situations was that they were, in
theory, quite mendable -- a program could easily have detected the incorrect
text-base files and queried the repository for the correct versions.  (This
is a consequence of the fact that the text-base is just a cache.)  Sometimes
I did this manually, and it worked as expected.

Eventually I stopped using the synchronized-WCs methodology, so I never got
motived to write the WC-cleaning script.

Dale


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

Re: Discontinuing the use of SVN.

[Ben Collins-Sussman <su...@collab.net>]

On Jun 3, 2005, at 4:34 PM, Robert Murphy wrote:

> After much frustration (and yelling) I have been told to remove SVN  
> from our production pipeline after the latest round of ‘Checksum  
> mismatch’ errors.

Sorry about that.  The checksums exist to point out hardware problems  
with network, disk, and memory.  I hope your hardware is okay!


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