[Top][All Lists]

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

Re: Do we have an alternative to tag, to record status of cvs repo

From: Paul Sander
Subject: Re: Do we have an alternative to tag, to record status of cvs repo
Date: Tue, 6 May 2003 13:15:21 -0700

>--- Forwarded mail from address@hidden

>On Tue, May 06, 2003 at 10:10:03AM -0700, Paul Sander wrote:
>> [...] my checkout
>> used RCS directly.  Bypassing the CVS directory-level locking protocol
>> is safe under these conditions, because the versions are known in advance
>> and are immutable.  The RCS file update algorithm and the filesystem
>> together avoid problems with lost data in the event that my checkout runs
>> concurrently with a commit.

>Did you guard against the critical section during a commit when the
>,v file doesn't exist, or did you find that in practice the
>window's too short to be a problem?

As far as I could tell, the rename system call used by RCS (when it replaces
the ,v file with the updated ",*," file) was atomic in our shop.  In any case,
I had never had a "file not found" error reported while opening an RCS file.
Such errors could always be treated as transient and be retried once or twice,
in any case.  An interesting corner case happens when an RCS file migrates
in or out of the Attic, but in practice that wasn't a problem, either.

>Also, did you manage to make your check-out-er work client/server?

A client/server implementation wasn't a requirement for that tool.  I see
no reason why it couldn't be done, though.

>--- End of forwarded message from address@hidden

reply via email to

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