Derek R. Price writes:
>
> I can actually come up with a broken-locks scenario to duplicate this
> corruption

That rings a bell.  There was a long-standing locking bug in tag where
we would collect some information about the RCS file with no lock in
place, then later we would take out a write lock, fully parse the file,
add the tag, and rewrite it.  I don't remember the details (if I ever
bothered to figure them out) of exactly what information was read when,
but it seems at least plausible that a commit between the initial read
and the full read would result in that kind of corruption.  The bug was
fixed quite a while ago, first by locking the whole tree for the
duration of the tag and then later by correctly locking one file at a
time, but if the broken revision is old enough, that might explain it.

-Larry Jones

It's going to be a long year. -- Calvin


_______________________________________________
Bug-cvs mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/bug-cvs

Reply via email to