Coda File System

Coda-6.9.7... 6.9.8 released

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Thu, 21 Jul 2016 16:04:55 -0400
I got ready to release Coda-6.9.7 at the end of May, but had some issues
with packaging on older Ubuntu systems. This then got interrupted by
vacation and by the time 6.9.7 was built and packaged (end of June) I
discovered that I had missed a couple of things.

So now we are about 2 months after I wanted to release, so I bumped
the version to "Coda-6.9.8" making it look like we are having frequent
monthly releases.

Changes since 6.9.6,

- Do not give default permissions to System:AnyUser on newly created
  volumes.
- Avoid a segfault in eprint (venus stdout/console log) where we
  traversed a variable argument list twice.
- Remove support for multiple Coda server processes on a single host where
  each was bound to a separate IP address using IP aliases on one interface.
- There was a buffer overflow in the Coda backup program, not super
  critical unless someone is backing up more than 252 Coda servers.
- Added systemd unit files. I used the ones from Fedora's Coda rpm
  packages, from when Fedora still had their own. Not sure how much
  systemd has evolved since then, but they seem to work.
- Make sure the uid value entered in vice-setup-user is a number.
- Avoid dumping binary crud to venus's stdout/console log.


Big question for 6.9.9,

Should we remove Coda's "higher level" backup tools (backup.sh, backup,
tape.pl, merge)?

We actually have not been using them for probably the past 15 years or
so and instead have successfully used Amanda, followed by BackupPC, and
finally settled on a homegrown solution (https://github.com/cmusatyalab/deltaic)
which uses thinly provisioned LVM volumes to maintain what is best
described as 'block-level deduplicated snapshots'.

Either way, unused code rots, any fixes for the rot end up not ever
getting sufficient testing and the names 'backup' and 'merge' are far
too generic for Coda specific programs.

Jan
Received on 2016-07-21 16:05:10