reported by Gary Duzan in PR pkg/22274
-being here, uodate to 5.3.20
changes:
- Updated for current autoconf/gcc build environment.
- Repair tries to continue even when not all replicas can be mounted.
- I _think_ this code fixes the problem where venus dies when it is
restarted after a local-global conflict and required reinitialization.
- Added 'masquerade_port' option to venus.conf to force venus to use a
fixed port when 'masquerade=1'. Simplifies life for strict firewall
administrators.
- Improved Kerberos intergration.
- Fixed the problem where servers would crash when more than 30000 files
were created in a single volume.
include:
* Installing the relevant manpages with the server programs.
* Miscellaneous bug fixes.
* Coda servers were doing strncmp instead of memcmp on ip-addresses.
This fixes 'identity-crisis' problems for replicated servers, where
servers didn't know which volumes they were hosting etc. (Ivan Popov)
* Made ValidateAttrs on the server more resilient to bad input.
* Handle reintegration conflicts when doing a partial reintegration,
instead of endlessly looping.
* Searching the colon separated CODACONFPATH environment variable for
venus.conf and server.conf configuration files.
* Added some more paranoia checks while parsing the /vice/db/servers
file.
* No longer ordering VRDB entries entries by ip-address. This allows
us to resize a VSG without affecting version-vectors.
* Added rds_print_stats to the server's printstats output to aid in
finding whether an RVM allocation failure is due to running out of
RVM space, or due to a large defragmentation that cannot be logged
in the RVM log.
* Added getvolinfo program to be useful when debugging volume mounting
problems.
* Avoid servers getting stuck in file-creation when there are
pre-existing files in /vicepa.
* Always pre-allocate fids with the maximum possible stride, will become
useful when we allowing modifications to a volume's replication
group. However, any replicated volume over a mix of old and new
servers has a slightly higher probability on getting unwanted
fid-collissions.
* Reduced server memory usage during reintegration.
* Not returning ENOENT when testing inconsistent files with
access(F_OK).
* Allow server-server repair with only one existing replica, needed
for repairing conflicts after running out of diskspace.
* Added "skip volumes" argument to norton-reinit -load.
* The 'change id' function of pdbtool would silently corrupt the pdb
databases, group owner id's were not updated, users were missing
as members from groups. Group could become a member of a user.
Hueristics to fix these corruptions are added to the 'pdbtool cm'
functionality.
