subversion-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bert Huijben" <>
Subject RE: Problem with adding files in SVN 1.8.0+. Is it in the tracker already?
Date Mon, 23 Dec 2013 14:07:22 GMT

> -----Original Message-----
> From: JANIKOVIC Jan []
> Sent: maandag 23 december 2013 12:58
> To: Bert Huijben; 'Geoff Field';
> Subject: RE: Problem with adding files in SVN 1.8.0+. Is it in the tracker
> already?
> Hello Bert,
> Would these reproduction steps help? If there is a way how to get a log
> or any other way to help fixing this issue, please let me know:
> Server installation: RHEL 4, Subversion svn, version 1.5.5
> Computer installation: TortoiseSVN 1.8.4, serf 1.3.2, computer restarted
> installation

I'm developing on Windows, so this makes it very hard to replicate this
problem for me...
Eventually your old report made it possible to replicate the HEAD problem
for me on Windows, which made me debug serf and Subversion.

And even then I would setup my repository using a pretty standard
configuration using a normal password backend; the default number of
requests, etc. etc. Your older reports noted that you didn't use a plain
text password backend, etc.

That is 100% essential information to get things reproduced for anybody.

Requiring a specific pretty old, platform to reproduce anything is making it
less likely that anybody can reproduce it.

Did you try your setup (=config files) on a setup that is actively supported
by any of the commercial vendors?

If you can that would really help.

> 1. Repository updated

How do you update a repository?

Let's assume

$ svnadmin create REPOSITORY

And hooking that to a url http://my-server/svn/repos

$ svn import greekfiles http://my-server/svn/repos/ -m ""
<answer username password, store in cache>
Adding         greekfiles\A
Adding         greekfiles\A\B
Adding         greekfiles\A\B\E
Adding         greekfiles\A\B\E\alpha
Adding         greekfiles\A\B\E\beta
Adding         greekfiles\A\B\F
Adding         greekfiles\A\B\lambda
Adding         greekfiles\A\C
Adding         greekfiles\A\D
Adding         greekfiles\A\D\G
Adding         greekfiles\A\D\G\pi
Adding         greekfiles\A\D\G\rho
Adding         greekfiles\A\D\G\tau
Adding         greekfiles\A\D\H
Adding         greekfiles\A\D\H\chi
Adding         greekfiles\A\D\H\omega
Adding         greekfiles\A\D\H\psi
Adding         greekfiles\A\D\gamma
Adding         greekfiles\A\mu
Adding         greekfiles\iota

Committed revision 1.
So now we have a repository...
(See our FAQ and 'HACKING' for some tricks to setup test environments)

> 2. File "new.txt" with arbitrary content copied to the repository folder
on the
> computer

Copying files to a repository directory is never recommended. Let's assume
you are talking about a working copy

$ svn co http://my-server/svn/repos wc
$ touch wc/new.txt
$ svn add wc/new.txt

> 3. Right-mouse click on the new file: TortoiseSVN->Add
> 4. Right-mouse click on the new file: SVN Commit
> 5. Press OK on the Commit dialog that appears

On this list we +- assume that you use 'svn', so let's assume that you did
$ svn commit -m "Message" wc
Adding wc\new.txt
Committed revision 2.

> 6. Form "Authentication" appears with following text:
> <server name> Authorization Realm
> Request username and password
> Username: [textbox]
> Password: [textbox]
> [checkbox] Save Authentication

This eventually documents that you did setup authentication on your
repository. We should add that information to step 1.
> 7. After third attempt to enter the login commit fails with following
> Error: Commit failed (details follow):
> Error: No more credentials or we tried too many times.
> Error: Authentication failed
> Error: Additional errors:
> Error: Error running context: The requested authentication type(s) are not
> supported

In your original report you didn't get authentication prompts here, so
something changed.

It would be very useful to add the updated server error logs here too. The
thing we try to solve is why these request fails and we now just know that
the authentication failed.

I don't know the exact details, but I remember something about Kerberos. Are
you used to seeing password prompts? (If you are using Kerberos/ntlm you
usually only see prompts when there is a problem connecting using the
default method)
Did you really type the right username and password (casing, prefix/vs no
prefix). Did some ticket expire?

The server log would be very informative here. (Like we explained in the
original report) as it usually has more details than what you do want to
send to your users.

The exact error message you note here is raised in our http layer after the
server reported the credentials as invalid 4 times. (The first might be an
attempt without a password; not sure). The original problem was that we
didn't authenticate a HEAD request at all. It would be good to know if this
works correctly now and if it is still the same request that fails.

> As for the tracker, adding the issue to it would help testers to see in
> version of serf, or svn client the bug was fixed and then I, or other
> parties, could test the fix when it will be added to TortoiseSVN. We could
> benefit also from the history in one location. Furthermore at the
beginning I
> spent some time to find the related discussion about this bug I believe
> other passive users of Tortoise SVN would find it easier to see that
> something is being done with this issue and that there is no workaround
> present yet apart from downgrading. That just how I see it.

As I see it adding an issue in this stage where an issue is reported by a
single party, without a way to reproduce it for anybody else is just a way
to postpone the problem. Maybe it is nice to see that it is still not
resolved with Subversion 1.10 and 1.11, but I'm more looking forward to
resolve the problem. If you just want to the state tracked for you I would
recommend asking one of the commercial parties backing the Subversion
development. (FYI: My work is indirectly paid from the income of that)

Other users can't really look at the existing issue as without a clear
reproducible description there is no way to know that their issue is really
the same issue as this one. The issue tracker is for know/identified issues
that we should be able to solve given enough time/resources and for feature
ideas that we may/may not implement later.

> Kind regards
> Jan
> -----Original Message-----
> From: Bert Huijben []
> Sent: Montag, 23. Dezember 2013 12:10
> To: JANIKOVIC Jan; 'Geoff Field';
> Cc: PETERS Michael
> Subject: RE: Problem with adding files in SVN 1.8.0+. Is it in the tracker
> already?
> > -----Original Message-----
> > From: JANIKOVIC Jan []
> > Sent: maandag 23 december 2013 11:52
> > To: Bert Huijben; 'Geoff Field';
> > Cc: PETERS Michael
> > Subject: RE: Problem with adding files in SVN 1.8.0+. Is it in the
> > tracker already?
> >
> > Hello Bert,
> >
> > Thank you for looking into this problem and for working on it. I
> > tested
> the file
> > addition in TortoiseSVN1.8.4 using serf 1.3.2 (released on Oct. 4,
> > after
> your
> > fix). The issue still exists there, but the behaviour is different
> compared to
> > TortoiseSVN 1.8.3: When user attempts to commit an added file, he is
> > prompted for login. Even when correct login is provided, the login
> > dialog appears again two more times after which the commit fails.
> > There is still
> no
> > tracker (TortoiseSVN, Subversion.Apache or serf) where this issue
> > would be tracked. Would it be possible to add it to one of these
> What would it help any of us to add it to a tracker?
> That doesn't magically solve this problem, does it?
> What we need is a good report of the problem that makes it possible to fix
> the problem. If we have that information we can fix it directly, or we
> (for different reasons) postpone fixing the problem. In that case it helps
> add it to a tracker.
> Just adding issues to a tracker just slows down fixing the actual problem.
> Issues require maintenance and it is not like we -as open source project-
> somebody to do that. And issues with not enough information to fix it will
> eventually (perhaps in a few years) just be closed as something like
> 'WORKSFORME' by a developer that takes the time to look through the issue
> tracker to see if there are things he can fix.
> Personally I'm quite easy to convince to fix an issue directly when
> hands me the information to reproduce the problem they see.
> If the issue is really important I'm even able to drop other work at hand
> to solve it. (Just compare the list traffic to the Subversion commits if
> need some examples :)). In this case an issue number is perhaps nice for
> changelog, but it doesn't really help either.
> The best bug reports are just a few simple steps that show how any
> developer can reproduce and debug the problem. (Well, perhaps patches
> are even better... but we can't expect the average user to debug through
> the low level network implementations)
> The information that the new serf changes the behavior is really
> but then you note that 'the commit fails'. There are at least
> 1000 different reasons why a commit can fail, so that last bit really
> help. Usually serf produces very cryptical, but for a developer very
> informative error messages and I would really recommend posting the errors
> you see here. (Or as noted a few times before: how we can see the errors
> for
> ourselves)
>         Bert
> ________________________________
> CONFIDENTIALITY : This e-mail and any attachments are confidential and may
> be privileged. If you are not a named recipient, please notify the sender
> immediately and do not disclose the contents to another person, use it for
> any purpose or store or copy the information in any medium.

View raw message