Milestone Coda-6.9.5

Completed 7 years ago (Mar 31, 2010, 2:42:25 PM)

100%

Total number of tickets: 6 - closed: 6 - active: 0

coda

4 / 4

coda/client

1 / 1

rvm

1 / 1

Coda-6.9.5, LWP-2.6, RPC2-2.10, RVM-1.17

  • When writing a checkpoint file of the reintegration log took longer than the checkpoint interval, we would immediately start writing out a new checkpoint, looping indefinitly. (Paolo Casanova)
  • Checkpointing failed when the reintegration log contained an empty file.
  • Truncate cache file when lookaside lookup fails to avoid the following fetch from assuming we already successfully fetched some of the data.
  • Make sure we wake up blocked threads when a fetch fails.
  • Only close the shadow file descriptor during reintegration if we actually opened it.
  • Return permission error when a user tries to rmdir a mountpoint.
  • Do not flush kernel caches whenever we check if a file is in use.
  • gcodacon improvements, notification rate limiting and window placement. (Benjamin Gilbert)
  • Reduce server->client RPC2 timeout from 60 to 30 seconds, reduces the time a client is blocked while callbacks are broken.
  • Introduce stricter locking on the server->client callback connections.
  • Make sure clients cannot indefinitely keep a callback break RPC busy preventing it from completing.
  • Write a stack backtrace to the log when an assertion fails.
  • Don't create /etc/modules.conf on newer Linux kernels (Adam Goode)
  • Don't install init scripts when building from source. Example scripts based on Hans de Goede's LSB scripts for Fedora are provided in tools/, but actually installing them is something that should now be done as a part of the distribution specific packaging.
  • Change the default venus.conf settings to use more standard locations for various log and cache files (/var/log/coda, /var/lib/coda). This changes only the default 'example' configuration file, so it will not affect settings for existing clients.
  • Avoid fortify error when we switch between LWP threads.
  • Make sure rpc2 always checks the global timeout during multirpc handling.
  • Send keepalives to RPC2 clients when the initial connection setup takes longer than expected. This avoids clients from disconnecting during connection setup when the server is busy.
  • Fixed various concurrency issues in RVM which were found while running with pthreads. It is unclear if these could occur with cooperative threading which is what Coda clients and servers use but it is recommended to update.
Note: See TracRoadmap for help on using the roadmap.