Re: TSVN update much slower than "svn update"

On 04.09.2009 17:42, Purple Streak wrote:
> 2009/8/12 Purple Streak<mrpurplestreak_at_googlemail.com>:
>> 2009/8/2 Stefan Küng<tortoisesvn_at_gmail.com>:
>>> On 31.07.2009 15:30, Purple Streak wrote:
>>>> 2009/7/23 Stefan Küng<tortoisesvn_at_gmail.com>:
>>>>> If it's the initial connection that takes much longer than in the svn
>>>>> command line client, then it's most likely an issue with SSPI
>>>>> authentication. The command line client doesn't have SSPI compiled in,
>>>>> so it won't have an issue with that at all.
>>>>>
>>>>> Make sure that your pc is properly joined to the domain server and that
>>>>> your IP address is properly set up and recognized in your DNS server. If
>>>>> one or both are not correct, SSPI takes a long time to find out that
>>>>> this kind of auth works/doesn't work.
>>>>
>>>> Hi Stefen
>>>> I've not been able to find anything wrong. It works fine from another
>>>> machine, so it's something on my current one. I've been tracking down
>>>> another issue with authentication and using cURL
>>>> (http://curl.haxx.se/) which can use sspi and that works instantly.
>>>>
>>>> Is there any way to get some logging out of TortoiseProc to see where
>>>> it's getting messages or should I get a debug build going to try it?
>>>
>>> you could try switching to serf instead of neon and see if that makes
>>> any difference. Other than that, there's no logging for such issues.
>>> Neon does have some logging, but that has to be compiled in and only
>>> works in a debug build. So if you really require that kind of logging,
>>> you'd have to build TSVN yourself and tweak the build scripts for neon
>>> to include that logging.
>>>
>>> To switch to serf, add the line
>>> http-library = serf
>>> in the svn servers config file (%APPDATA%\Subversion\servers)
>>>
>>
>> Hi - just got back to looking at this again. Switching to serf does
>> make a big difference. However serf seems to make the command line
>> slower (and my perception is that TSVN is still slower then it used to
>> be and slower that it is for my collegues). For a simple svn list it
>> almost doubles from 0.3s to 0.6s. Not that this is a lot, but when
>> my TSVN was fast then it made the repo browser so much more
>> responsive. Over a largeish checkout serf takes 3m41s, neon takes
>> 2m43s
>>
>> You also said:
>>>>> The command line client doesn't have SSPI compiled in,
>>>>> so it won't have an issue with that at all.
>>
>> From what i've looked at the command line uses neon to do SSPI. If I
>> tunnel the connection through a proxy the command line seems to be
>> using only NTLM auth, and never asks me for a password (despite having
>> cleared my auth cache) where as TSVN does sometimes ask for this.
>> Does TSVN have it's own SSPI code in somewhere? Otherwise I can't see
>> why it would work fine for the command line but not for TSVN?
>
> Hi - Sorry to bump this but I'm still stuck looking for where this
> would be different in TSVN (see my previous query about)? I've got a
> debug tortoise built, so if you could point me at the bit where
> tortoise does the SSPI then I can try and debug what is happening.

If TSVN asks for authentication only sometimes, it's usually because the
diff to transfer is too big (maybe that's why you haven't seen this with
the CL client, you might have not done the same big commits/updates/...
as with TSVN).
See here:
http://svn.haxx.se/tsvnusers/archive-2008-02/0329.shtml

Stefan

-- 
        ___
   oo  // \\      "De Chelonian Mobile"
  (_,\/ \_/ \     TortoiseSVN
    \ \_/_\_/>    The coolest Interface to (Sub)Version Control
    /_/   \_\     http://tortoisesvn.net
------------------------------------------------------
http://tortoisesvn.tigris.org/ds/viewMessage.do?dsForumId=4061&dsMessageId=2391049
To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_tortoisesvn.tigris.org].