1. 06 Apr, 2014 2 commits
    • cjmueller's avatar
      Updated debian/changelog · e7acbf10
      cjmueller authored
    • cjmueller's avatar
      Version 1.05 · 38cc2000
      cjmueller authored
      - Allow SCSI device names with more than one character (e.g. sdaa) in case
        there are more than 26 SCSI targets.
  2. 26 Oct, 2011 3 commits
  3. 05 Dec, 2010 1 commit
    • cjmueller's avatar
      Version 1.03 · 388462c3
      cjmueller authored
      - Use %u in dprintf() when reporting number of reads and writes (the
        corresponding variable is an unsigned int).
      - Fix example in README where the parameter "-a" was written as "-n".
  4. 06 Nov, 2010 2 commits
  5. 26 Feb, 2010 1 commit
    • cjmueller's avatar
      Version 1.01 · 388ef4d0
      cjmueller authored
      - The parameter "-a" now also supports symlinks for disk names. Thus, disks
        can be specified using something like /dev/disk/by-uuid/... Use "-d" to
        verify that the resulting disk name is what you want.
        Please note that disk names are resolved to device nodes at startup. Also,
        since many entries in /dev/disk/by-xxx are actually partitions, partition
        numbers are automatically removed from the resulting device node.
      - Not really a bug, but the disk name comparison used strstr which is a bit
        useless because only disks starting with "sd" and a single letter after
        that are currently considered. Replaced the comparison with strcmp()
  6. 18 Nov, 2009 2 commits
    • cjmueller's avatar
      *** empty log message *** · 1f1b4e65
      cjmueller authored
    • cjmueller's avatar
      Features · 04db2050
      cjmueller authored
      - New parameter "-a" to allow selecting idle timeouts for individual disks;
        compatibility to previous releases is maintained by having an implicit
        default which matches all SCSI disks
      - Changed comparison operator for idle periods from '>' to '>=' to prevent
        adding one polling interval to idle time
      - Changed sleep time before calling sync after updating the log file to 1s
        (from 3s) to accumulate fewer dirty blocks before synching. It's still
        a compromize but the log file is for debugging purposes, anyway. A test
        with fsync() was unsuccessful because the next bdflush-initiated sync
        still caused spin-ups.
  7. 24 Apr, 2007 3 commits
  8. 23 Apr, 2007 1 commit