Re: how to prevent cvs from locking?

[DC]

On Thu, 22 Sep 2005 15:47:29 -0400, Jim Hyslop <address@hidden> wrote:

>DC wrote:
>> I would rather put a non-existant email address out there, then one that im 
>> never going to look at. At least they might get a bounce back if
>> it dosn't exist.  But, for the sake of newsgroup postings, I have created a 
>> gmail account.
>[snicker] Nice name for the spam trap.
>
>> it is a very small group of developers. Most of the time they will be in 
>> completely different places on the branch.
>Doesn't matter. All it takes is for someone to do a repository-wide 
>command, such as applying a tag, at the same time as someone else is 
>trying to write to the repository and your repository will become 
>corrupted, possibly beyond repair.
>
> > For the most part people
>> don't touch each others files, and when they do there is usually a 
>> discussion of how/what/what/etc before hand.
>Ah, the strong, strong smell of code ownership.
>
>> If it were my descision I would set this up so that people create files as 
>> username:cvspub rather then cvspub:cvspub.
>That's reasonable.
>
>> I would also make it
>> so that two people could not checkout the same file at the same time.
>That's not ;=) Remember that the "C" in "CVS" stands for "Concurrent" - 
>CVS is designed specifically to allow and handle concurrent 
>modifications to files. But that's another rant for another day, and 
>doesn't solve your immediate problem, which is:
agreed :)


>
>> Error:
>> cvs: rcs.c:4114: RCS_checkout: Assertion `rev == ((void *)0) || 
>> ((*__ctype_b_loc ())[(int) (((unsigned char) *rev))] & (unsigned short int)
>> _ISdigit)' failed. 
>> Terminated with fatal signal 11 
>
>Because CVS gets this fatal signal, it doesn't get a chance to clean up 
>after itself. The lock files are left behind, causing problems for the 
>next person who comes along.
>
>Larry Jones has suggested this error means your repository file is 
>corrupted. If other users try to check out the file, do they get the 
>same crash?
>
>Download the source code from www.nongnu.org/cvs, and run the 
>validate_repo script in the contrib/ directory. This will scan the 
>repository for corrupted files. If the file is corrupted, the easiest 
>course of action is to restore it from your most recent backup. Also, 
>you must try to determine how it got corrupted in the first place, 
>otherwise this problem could happen again.


Im trying the validate_repo script now... before I did that, i went on my 
computer and checked out the HEAD branch without any trouble what
so ever.