[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: cvs repository files being corrupted

From: Todd Denniston
Subject: Re: cvs repository files being corrupted
Date: Mon, 13 Feb 2006 18:09:31 -0500
User-agent: Mozilla Thunderbird 1.0.7-1.1.fc4 (X11/20050929)

Please keep the replies on list as others may be able to use the info to either help you or figure their problem out from the archives.
Jeremy Birch wrote:

Q1) is your repository on an nfs or smb share (or other network filsystem)?

no - it is local to the server

Q2) what access method are you using, i.e., :ext:, :pserver:, :local:, or direct to drive?


Q3) which distribution version of Linux are you using?

SuSe 9.1

This reminded me of some problems seen back in 2001:


I think validate_repo.pl is something you may want to use as well, it searches for faults with the ,v files IIRC. http://cvs.savannah.nongnu.org/viewcvs/ccvs/contrib/validate_repo.pl?rev=1.1&root=cvs&view=log

some of these bugs seem similar expect our is ALWAYS just after the "desc @@" in the file, no matter which file it is. It tends to happen more on big binary files so we suspect it has something to do with how long that file is being worked on (ie higher probability of a clash with some other process) but have no confirmation of this.

We are working in client-server mode, no one is using NFS mounts or other such mechanisms to access the repository.


BTW which client mode, it may make a difference in later debugging.

It seems to happen most often on tags (but of course that tickles a lot of files) but we have seen it on individual ascii file commits too

Any ideas and suggestions gratefully accepted

1) What kind of access to the physical hard drive is your server using?
(if the server is accessing the repository over NFS it can cause these faults, it is not necessarily a client thing.

2) How old is the drive/cpu/memory on the computer?
it might be time to do a good (well as good as is possible with already faulting parts) backup, and run badblocks (with full write testing, which is why you HAVE TO HAVE THE BACKUP) and memtest86 on the system.

3) this might be a really long shot, and I would not expect it of a modern linux system, but you might try turning off the memmap option in the compile. [search bug-cvs for where I had to do it to an old sun machine, to see how.]. if the other things show no problems and the memmap change does "fix" the problem please let us know.



Todd Denniston
Crane Division, Naval Surface Warfare Center (NSWC Crane)
Harnessing the Power of Technology for the Warfighter

reply via email to

[Prev in Thread] Current Thread [Next in Thread]