pcp
[Top] [All Lists]

pcp updates: libpcp, qa

To: pcp developers <pcp@xxxxxxxxxxx>
Subject: pcp updates: libpcp, qa
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Fri, 4 Sep 2015 03:09:55 -0400 (EDT)
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <2118363591.26687827.1441350585687.JavaMail.zimbra@xxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: zMp0yhWx2LiLEprO1gUUEGhLTrXlwA==
Thread-topic: pcp updates: libpcp, qa
Changes committed to git://git.pcp.io/pcp.git master

Nathan Scott (2):
      qa: add diagnostics to 785 to help figure out a buildbot failure
      libpcp: optimise archive decompression for common operations


 qa/785                   |    1 
 src/libpcp/src/logutil.c |   98 ++++++++++++++++++++++++++++-------------------
 2 files changed, 61 insertions(+), 38 deletions(-)


commit 12da325fa449ad775b892437e84b8db0ef60fbac
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Fri Sep 4 17:07:10 2015 +1000

    libpcp: optimise archive decompression for common operations
    
    While testing pcp-atopsar(1) it was noticed to be invoking the
    uncompression program far more than necessary.  On inspection,
    found the libpcp log decompression caching being circumvented
    deep in pmFetch - due to __pmLogSetTime calling _logpeek when
    performing its truncation checking on the last volume.  This
    then dodges the loglabelchk optimisations (to not reopen logs)
    accidentally and we end up in a repeated decompress/unlink tmp
    file cycle in the fetch loop.
    
    Also made code legibility changes to make some parts easier to
    follow and made the current default compression program/suffix
    the first one looked up, rather than the last.

commit 603569fb19e92a79db002d31f9885e6b770506d3
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Fri Sep 4 12:20:01 2015 +1000

    qa: add diagnostics to 785 to help figure out a buildbot failure


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