[cvsnt] file corruption using CVSNT against non-CVSNT

Paul Balmforth pab_online at yahoo.com
Fri Aug 18 17:21:46 BST 2006


Community technical support mailing list was retired 2010 and replaced with a professional technical support team. For assistance please contact: Pre-sales Technical support via email to sales@march-hare.com.


>> - it places a responsibility on the client to parse and evaluate the
>> wildcards described in cvswrappers (or wrapper options). The new
> 
> No it doesn't - no client does that - in fact we recommend disabling the 
> ability of clients to override the wrappers on the server since clients 
> often have different ideas of what is binary.
   
  How does the client determine arguments to the 'Kopt' request at import time, before files have ever been checked out? It must parse and evaluate wrapper wildcards against filenames.
   
  Since I'm developing a pure client rather than client/server in one, I'm acutely aware of responsibilities on clients which could be handled by the server. The whole idea of the 'Kopt' request seems somewhat flawed, because determination of the kopt can (or should) always be done by the server. I'm guessing that the decision to allow wrappers to be defined in an environment variable or ~/.cvswrappers was the basis for the CVS team's decision to add 'Kopt' to the protocol, because the server knows nothing about them. The fact that my client doesn't support the 'Kopt' request has never been much of a problem, since the server (whether CVSNT or Cyclic) has always consolidated CVSROOT/cvswrappers and the -W options received through Argument requests and used them to determine the kopt of the file.
   
  > Tested and cvsnt is behaving correctly - the file is not corrupted and 
> is correctly stored as binary.  compatibility with cvs 1.11 is one of 
> those things that gets a lot of testing through various means and 
> breakage like that would have been noticed quickly.
   
  I have two CVSNT servers set up here. One is 2.5.02.2088 (Servalan), the other is 2.5.03.2151 (Scorpio). The problem reproduces for me against Servalan but not Scorpio. However, we have a customer who is complaining of the same issue in Scorpio and even 2.5.03.2260, so this seems to relate to the configuration or environment rather than the build number.
   
  Having extended my client for the 'Kopt' request I'm performing two identical test runs of import and changing only one parameter - the argument to the 'Kopt' request. If passing 'Kopt b' everything works nicely, but 'Kopt -kb' leaves the file corrupted.
   
  > Writing a cvs client from scratch is a nontrivial task - it's very hard 
> to get all the interaction right so that the server knows exactly what 
> to do... I would hesitate to do it myself and I've been working with the 
> protocol for 5 years.
   
  Our product also supports interoperability with an external CVS executable. The client we have developed will never be as feature-rich as CVSNT, and it's not our intention to make it so. It's already 3 years old, basically functional, and stable.
   
  Paul.
   

 		
---------------------------------
 Now you can scan emails quickly with a reading pane. Get the new Yahoo! Mail.


More information about the cvsnt mailing list
Download the latest CVSNT, TortosieCVS, WinCVS etc. for Windows 8 etc.
@CVSNT on Twitter   CVSNT on Facebook