Folks, this weekend has been a rough one for Kamesh and myself. We wrangled
with a CollabNet BDB repository corruption that turned out to be
Subversion's fault, causing a loooot of overtime work hours and (at least in
my case) a lot of lost sleep. So please allow me the luxury of *not*
repeating the whole story. The parts you care about are at:
http://subversion.tigris.org/issues/show_bug.cgi?id=2751
For those seeking drama, this was my first "opportunity" to actually be
doing more-or-less manual manipulation of individual BDB repository table
rows, breaking the cycles and restoring the repository into working
conditional by hand. What an adventure!
Anyway, I'm mailing the list because I have a couple of needs. First, I'd
really appreciate peer review of the fix itself. It's straightforward
enough, passes 'make check', has a regression test, and I've done the math
plenty of times. But having another fs-knowledgable developer say, "Yep,
looks great" would be very comforting.
Secondly, in the event that the fix is deemed right, I'd like to propose
that we escalate the release of the fix as follow-up revisions to the 1.4.x
and 1.3.x lines. And it wouldn't break my heart to patch all the way back
to 1.0.x, especially as I'd expect folks still on those revisions to be
using the BDB backend, and with the exception of a one-time minor rewrite of
the merge() function, I suspect that area of code hasn't really changed all
that much since pre-1.0.
Thanks.
-- C-Mike
PS: Despite the circumstances, I really had a blast diving that deeply into
Subversion's code again. I hope to free up more time to do so in coming months.
--
C. Michael Pilato <cmpilato@collab.net>
CollabNet <> www.collab.net <> Distributed Development On Demand
Received on Mon Mar 26 05:39:22 2007