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.
Gates, Donald T wrote: > > Hmm. Well, since I AM running Symantec Anti-Virus Corporate Edition v.7x, > I could test it. What was your testing methodology? Do you have a test > script(s) that I can execute/perform? > >Basically when the error kicks in any new file added to the repository will >have a corrupt diff entry for each new version committed. It seems once it >starts happening it happens 100% of the time. > >Since it's only repeatable on machines running NAV and doesn't seem to >repeat using the free trial edition I've resorted to going through the code >and saying 'what if <x> went wrong'. Every single fopen/rename/read is >error checked as far as I can see, so you'll get a trace if there's an FS >error... the only conclusion from the data that diff is generating is that >the file isn't returning an error but is returning incorrect data. >Basically CVS writes out two temp files, which are the old and new state of >the file, then calls 'diff' on them to generate the deltas for entering >into the RCS file. Either the write or the subsequent read isn't doing >what it should do. > >The best way to test would be to first trigger it, then disable stuff until >the repository starts behaving sanely again. I'd start with the 'worst' >situation (NAV fully enabled on both client and server) as that seems most >likely to involve the trigger condition. > >Of course the low number of reports of this tend to suggest it's not as >simple as 'Install NAV and it breaks'. There could be other things >interacting as well. > >This is what one person did to repeat it: >> >>echo a line >newfile >> >>cvs server: scheduling file `newfile' for addition >>cvs server: use 'cvs commit' to add this file permanently >> >>cvs ci >cvs commit: Examining . >RCS file: C:/Docume~2/cvs/MyDocu~1/cvsroot/website/newfile,v >done >Checking in newfile; >C:/Docume~2/cvs/MyDocu~1/cvsroot/website/newfile,v <-- newfile >initial revision: 1.1 >done > >>echo another line >>newfile > >>cvs ci -m"Bye Bye" >cvs commit: Examining . >Checking in newfile; >C:/Docume~2/cvs/MyDocu~1/cvsroot/website/newfile,v <-- newfile >new revision: 1.2; previous revision: 1.1 >done > > >... you can use extra '-t' arguments to cvs for extra verbosity, so when/if >you trigger it then a verbose trace would tell me quite a bit to see >exactly where it's failing and put in a fatal abort if I can detect it. > >Tony I'm going on vacation (to the Bahamas no less!) for my 40th birthday, and won't be back until mid-January, but I will definitely give this a try when I get back and get my results back to everyone.