pcp
[Top] [All Lists]

pcp updates: qa

To: pcp@xxxxxxxxxxx
Subject: pcp updates: qa
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Tue, 15 Oct 2013 22:43:04 -0400 (EDT)
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <62317057.2654480.1381891349629.JavaMail.root@xxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: p+f2cxFc2ZIi6w2MJHrV6vfGWrzvMQ==
Thread-topic: pcp updates: qa
Changes committed to git://oss.sgi.com/pcp/pcp.git dev

 qa/067                     |    9 +++++----
 qa/067.out.1               |    2 +-
 qa/067.out.2               |    2 +-
 qa/067.out.3               |    2 +-
 qa/067.out.4               |    2 +-
 qa/749                     |   23 +++++++++--------------
 src/libpcp/src/auxserver.c |    6 ++++--
 src/libpcp/src/derive.c    |   17 +++++++++++++++++
 8 files changed, 39 insertions(+), 24 deletions(-)

commit 8da1d1a68902b3ee27110143c3fd4b7daf741676
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Wed Oct 16 13:39:47 2013 +1100

    Fix a race condition in test qa/749 (pmcd probes)
    
    Using advice from fche, rework the stap script in this
    test to run the PDU-generators (probe hitters) within
    a begin{} probe, to remove races around ensuring the
    pmcd probe handler is successfully activated.  Before,
    we were sometimes running the PDU-generators before we
    had stap fully setup, and hence events were missed and
    ultimately the test failed.
    
    As an added bonus, the test runs more quickly and is a
    fair bit simpler (no backgrounding of stap needed).

commit b41ff31b3f1bbd424ec7e130cc4da0b3b49c022a
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Wed Oct 16 13:34:26 2013 +1100

    Handle absense of server unix domain socket file gracefully
    
    No point warning that we couldn't remove the file during the
    server process shutdown if something has already removed it.
    Also, switch to using thread safe message reporting routines
    and platform-agnostic errno manipulation.

commit c69d62dcb2efd5adf1a320f45656c25e768c8678
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Tue Oct 15 14:46:47 2013 +1100

    Address potential lock inversion issues in libpcp derived metrics
    
    Fix Coverity-reported lock inversions in derived metrics.  However,
    these all look benign ... they result from calling __pmNoMem which
    means the game is all over (we might deadlock before exiting in the
    worst case).

commit 21be6171df53b59ccf4bfd70e204b675e290be34
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Tue Oct 15 14:35:22 2013 +1100

    Resolve recent failures observed in test qa/067
    
    A combination of factors has resulted in 067 beginning to
    fail recently - (re)introduction of a proc PMDA with domain
    number 3 (used by "test" PMDA in 067), and logging pmdaproc
    metrics like proc.nprocs, proc.runq and friends.
    
    Address this through use of a more appropriate domain number
    for "test" (using "lab" from old tutorials) which prevents
    pmcd from sending client requests through to the fake PMDA.

<Prev in Thread] Current Thread [Next in Thread>
  • pcp updates: qa, Nathan Scott <=