You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Jon Wells <jo...@bigbang.com.au> on 2004/06/20 01:37:46 UTC
subversion is broken
Anything which lets the repository become inaccessible because the
connection to the client fails is just WRONG.
Give up, go do something useful like watch TV.
jon.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion is broken
Posted by Ben Reser <be...@reser.org>.
On Mon, Jun 21, 2004 at 10:00:19AM +1000, Jon Wells wrote:
> I thought I was reasonable specific.... if the server is incapable of
> surviving a failed transaction with the client... which it is.. then
> subversion is useless in a distributed development group.
>
> The problem is two fold...
> (a) the server can get messed by the client failing to complete cleanly.
> (b) this can the entire repository to become inaccessible to all others.
>
> It doesn't matter what the cause of (a) is, it happens.
>
> CVS, which we've reverted to using, fails point (a) but the damage caused is
> very isolated.
>
> (b) is a show stopper.
I think what Brane was trying to say, though perhaps not in the most
diplomatic way (not that you were being diplomatic either), is that:
a) We'd like to fix these problems. We're trying to do so as we become
aware of places where these problems are caused.
b) We can't do that without more information. If you have information
that can help us find a problem like this or maybe even something like a
replication recipe, we'd really like to have it.
Unfortunately berkeley db doesn't really like programs shutting down
unexpectedly. Obviously, we try to prevent that from happening. But
not only that some effort has gone into making a file system level
implementation that doesn't use berkely db.
The ironic thing of all this is that whlie we're trying to fix all these
things and are asking for your help in doing so, you tell us to
"Give up, go do something useful like watch TV." Yet, you go back to
using a piece of software that you claim has a similar problem, but
just isn't as bad in your opinion.
Don't be surprised when you send inflamatory emails when you get
inflamatory replies back.
--
Ben Reser <be...@reser.org>
http://ben.reser.org
"Conscience is the inner voice which warns us somebody may be looking."
- H.L. Mencken
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion is broken
Posted by kf...@collab.net.
Jon Wells <jo...@bigbang.com.au> writes:
> Abort the client by hand in the middle of a long commit... using
> mod_dav_svn as the access mechanism... happened to more than one person
> in one week.
>
> Something somewhere crashed once while being executed by apache... I
> don't know what, I'm not going to look again, I don't care, I don't use
> it any more. But that also lead to the repository being screwed and my
> giving up on it. Server host's was an OpenBSD box, clients Open or Lunux
Can you tell us what version of Subversion (client and server) and
Apache HTTPD was used, and what the commit command looked like? How
many files were involved, how big were they?
(The file http://svn.collab.net/repos/svn/trunk/BUGS has more specific
guidelines on reporting bugs.)
If we can reproduce this, we can fix it. However, we can't reproduce
it right now. I've interrupted commits over ra_dav before, it's never
been a problem. If we learn your special circumstances, that may help
us track the bug down.
Now for the annoyed portion of the mail:
You wrote "I thought I was reasonable specific...." [sic]. Yet your
first mail did not even say *which operation* you were performing that
caused the repository to fail. Never mind advanced esoterica like,
say, the version of Subversion or the access method.
Your mail wasn't reasonably specific, by any stretch of the
imagination. It was uselessly, lazily vague, and you shouldn't be
surprised that it got no useful response.
Good luck with CVS. You make me glad I unsubscribed from their
bug-reporting list.
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion is broken
Posted by Ben Reser <be...@reser.org>.
On Mon, Jun 21, 2004 at 12:10:43PM +1000, Jon Wells wrote:
> Something somewhere crashed once while being executed by apache... I
> don't know what, I'm not going to look again, I don't care, I don't use
> it any more. But that also lead to the repository being screwed and my
> giving up on it. Server host's was an OpenBSD box, clients Open or Lunux
If you didn't care why'd you bother to send us an email?
--
Ben Reser <be...@reser.org>
http://ben.reser.org
"Conscience is the inner voice which warns us somebody may be looking."
- H.L. Mencken
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion is broken
Posted by Jon Wells <jo...@bigbang.com.au>.
Abort the client by hand in the middle of a long commit... using
mod_dav_svn as the access mechanism... happened to more than one person
in one week.
Something somewhere crashed once while being executed by apache... I
don't know what, I'm not going to look again, I don't care, I don't use
it any more. But that also lead to the repository being screwed and my
giving up on it. Server host's was an OpenBSD box, clients Open or Lunux
jon.
On Mon, 2004-06-21 at 11:59, Greg Hudson wrote:
> On Sun, 2004-06-20 at 20:00, Jon Wells wrote:
> > I thought I was reasonable specific.... if the server is incapable of
> > surviving a failed transaction with the client... which it is.. then
> > subversion is useless in a distributed development group.
>
> But we're not aware of any bug of this nature, and you haven't even
> described why you think the client dropping the connection is
> responsible for any repository inaccessibility you're seeing, much less
> any information which could help us track down such a bug if you were
> seeing it.
>
> If Subversion were useless in a distributed development group, we
> wouldn't be able to use it for, say, the Subversion project.
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion is broken
Posted by Greg Hudson <gh...@MIT.EDU>.
On Sun, 2004-06-20 at 20:00, Jon Wells wrote:
> I thought I was reasonable specific.... if the server is incapable of
> surviving a failed transaction with the client... which it is.. then
> subversion is useless in a distributed development group.
But we're not aware of any bug of this nature, and you haven't even
described why you think the client dropping the connection is
responsible for any repository inaccessibility you're seeing, much less
any information which could help us track down such a bug if you were
seeing it.
If Subversion were useless in a distributed development group, we
wouldn't be able to use it for, say, the Subversion project.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion is broken
Posted by Jon Wells <jo...@bigbang.com.au>.
I thought I was reasonable specific.... if the server is incapable of
surviving a failed transaction with the client... which it is.. then
subversion is useless in a distributed development group.
The problem is two fold...
(a) the server can get messed by the client failing to complete cleanly.
(b) this can the entire repository to become inaccessible to all others.
It doesn't matter what the cause of (a) is, it happens.
CVS, which we've reverted to using, fails point (a) but the damage caused is
very isolated.
(b) is a show stopper.
jon.
----- Original Message -----
From: "Branko ÄOibej" <br...@xbc.nu>
To: <jo...@bigbang.com.au>
Cc: <de...@subversion.tigris.org>
Sent: Sunday, June 20, 2004 6:24 PM
Subject: Re: subversion is broken
> Jon Wells wrote:
>
> >Anything which lets the repository become inaccessible because the
> >connection to the client fails is just WRONG.
> >
> >Give up, go do something useful like watch TV.
> >
> >
> If you have a specific problem, describe it and we can try to fix it.
>
> If, on the other hand, all you're capable of is spouting abuse, then
> kindly go do it somewhere else.
>
> -- Brane
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion is broken
Posted by Branko Čibej <br...@xbc.nu>.
Jon Wells wrote:
>Anything which lets the repository become inaccessible because the
>connection to the client fails is just WRONG.
>
>Give up, go do something useful like watch TV.
>
>
If you have a specific problem, describe it and we can try to fix it.
If, on the other hand, all you're capable of is spouting abuse, then
kindly go do it somewhere else.
-- Brane
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org