pcp
[Top] [All Lists]

pcp updates

To: pcp@xxxxxxxxxxx
Subject: pcp updates
From: Nathan Scott <nscott@xxxxxxxxxx>
Date: Fri, 12 Jun 2009 15:11:52 +1000 (EST)
Changes committed to git://oss.sgi.com/nathans/pcp.git

 man/man3/mmv_inc_value.3       |    2 -
 man/man4/mmv.4                 |   18 +++++-----
 src/include/mmv_stats.h        |   68 ++++++++++++++++++++---------------------
 src/libpcp_mmv/src/mmv_stats.c |    2 -                                        
 src/pmdas/mmv/mmv.c            |   64 ++++++++++++++++++--------------------   
 src/pmdas/windows/pmda.c       |    2 -                                        
 6 files changed, 76 insertions(+), 80 deletions(-)                             

commit 66194b20e1984f08eecc6163a14d3c6cb4391e41
Author: Nathan Scott <nathans@xxxxxxxxxx>      
Date:   Fri Jun 12 15:04:15 2009 +1000         

    hinv.ndisk should have discrete, not instantaneous semantics in Windows PMDA

commit c5c5740e919e1b3c852269e2b140ee85864103a7
Author: Nathan Scott <nathans@xxxxxxxxxx>      
Date:   Fri Jun 12 15:03:17 2009 +1000         

    MMV agent and library fixes, big and small.

    - fix man page cut&paste typo on mmv_inc_value NAME section
    - use double-backquote in mmv.4 section describing MMV magic
    - document the now-fixed sizes of MMV file format fields, esp.
      TOC struct which was not correct for 64-bit platforms (time_t).
    - fix mmv.4 typos - Lenght and secion, cut&pasted several times.

    - use fixed size types in the on-disk format for MMV mmapped file,
      this change largely cosmetic except for removal of time_t.
    - fix up some minor whitespace crimes in mmv_stats.h

    - pmdammv had a 32 bit (int) value for gen number in-core, but it
      was time_t on disk (which is a long - 64 bits for 64 bit hosts).
    - if HAVE_MKSTEMP was defined, we botch it completely in two ways:
      o  path[] buffer used to hold both targets of rename(2), so we
         attempt to rename a file to itself.
      o  free() of the filename was only correct for the tempnam usage,
         which mallocs the filename.  Currently we double-free at the
         end of map_stats which glibc detects and implodes.
    - remove a couple of unneeded, shadow declarations of "int i" in
      the map_stats routine.
    - we should not unlink the temporary namespace file explicitly, as
      this is now done in the ReplacePmnsNames script.

<Prev in Thread] Current Thread [Next in Thread>