pcp
[Top] [All Lists]

pcp updates: qa, build tweaks

To: PCP Mailing List <pcp@xxxxxxxxxxx>
Subject: pcp updates: qa, build tweaks
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Tue, 15 Jul 2014 22:44:51 -0400 (EDT)
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <58674295.10958187.1405475932087.JavaMail.zimbra@xxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: SYUIIotUbv2NepJjIOweTTKph31PHQ==
Thread-topic: pcp updates: qa, build tweaks
Changes committed to git://git.performancecopilot.org/pcp.git dev

 CHANGELOG                      |   25 +++++++++++++++-
 build/rpm/fedora.spec          |    2 -
 debian/changelog               |    4 +-
 qa/635                         |   25 +++++++++-------
 qa/775                         |   42 +++-----------------------
 qa/946                         |   42 +++-----------------------
 qa/GNUmakefile                 |    3 +
 qa/common.discovery            |   64 +++++++++++++++++++++++++++++++++++++++++
 src/libpcp/src/probe.c         |    6 +++
 src/pmfind/pmfind.c            |   22 ++++++--------
 src/pmieconf/GNUmakefile       |   21 +++++++------
 src/pmieconf/GNUmakefile.rules |    2 -
 12 files changed, 146 insertions(+), 112 deletions(-)

commit 78c82b24b5cb1072245bdd69284ab0e042179880
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Wed Jul 16 11:39:45 2014 +1000

    Fix test qa/635 for when netstat RX-OVR or TX-OVR are non-zero
    
    Experience has shown that the OVR fields are not as was thought
    for the original version of this test, i.e. just part of the
    drops metric count (and missing from /proc/net/dev).  Turns out
    the fifo column from procfs is what ends up being reported for
    the netstat OVR columns.
    
    This commit simply separates out the fifo metric handling in the
    test from the drop metric (were summed, previously) and verifies
    the two metrics independently against the two netstat columns.

commit df04118e59c0d5a5f178e0109b7606fda10adbeb
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Wed Jul 16 11:26:18 2014 +1000

    pmfind: fix the build on Solaris, without sighandler_t type

commit a3fc135bf1a04bd8fe6d2fc1d66cf770e923d3d8
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Wed Jul 16 11:07:15 2014 +1000

    Update changelogs and so on, prep for pcp-3.9.7 release

commit e42a6be87af2289b8cd99db87360a1e2da7b5c4c
Author: Svante Signell <svante.signell@xxxxxxxxx>
Date:   Wed Jul 16 10:51:48 2014 +1000

    Resolve pthread stack size build issue on GNU/Hurd platform
    
    On Hurd there is no pthread minimum stack size defined.
    On GNU/Linux this value is set to 16384.  We use that number
    as default when PTHREAD_STACK_MIN is not defined.  In fact
    for Hurd it could be as small as 4096, the page size for x86.

commit e8e889bdfa3201161b3923b975fa004399d8f902
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Wed Jul 16 10:48:18 2014 +1000

    Attempt to workaround a GNU/Hurd build issue for Svante
    
    Problem report is as follows, the pmieconf make rules are all
    a bit wierd, with the subdir descent making things awkward.
    
    From: Svante Signell <svante.signell@xxxxxxxxx>
    "Due to some yet unknown reason the second time pmieconf is creating
    the file 'local' it fails with the fname being empty in rules.c. The
    attached patch makefile.patch is a work-around for this problem by not
    deleting that file the second time.
    
    (cd src/pmieconf; PCP_CONF=../../src/include/pcp.conf
    
LD_LIBRARY_PATH=../../src/libpcp/src:../../src/libpcp_pmda/src:$LD_LIBRARY_PATH
    HOME=`pwd` PCP_ALT_CPP=../../src/pmcpp/pmcpp ./pmieconf -F -r rulesdir
    -f local) manually works to create 'local', but not in the nested
    make calls.
    
    It seems that the re-creation of that file could be avoided by
    modifying the files GNUmakefile and/or GNUmakefile.rules resulting in:
    Nothing to be done for 'default_pcp' when getting to the examples
    directory? (Dunno how to do that yet)."
    
    For now, I've rearranged the targets a little to try to avoid
    the problem and included parts of Svante's approach too, but the
    dependencies here are still not quite right and several things
    are always rebuilt when they shouldn't be (with or without this
    change, that is).

commit 3aa351852bfd94634cf4a9dae43c36f15f87afb9
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Wed Jul 16 09:15:01 2014 +1000

    Add common.discovery file into the pcp-testsuite installed set

commit e90d27ffb137a305c74c98e53456f21308b8a73c
Merge: ceb0640 07423d8
Author: Dave Brolley <brolley@xxxxxxxxxx>
Date:   Tue Jul 15 15:52:57 2014 -0400

    Merge remote-tracking branch 'origin/dev' into brolley/dev

commit ceb06409cbbf4b7942c26713b1b7ac823bfa4810
Author: Dave Brolley <brolley@xxxxxxxxxx>
Date:   Tue Jul 15 15:47:38 2014 -0400

    Add a filter for ignoring unresolvable addresses for service discovery 
tests.
    
    The new file qa/common.discovery contains a new filter,
    _filter_discovery_unresolvable and moves other
    common code from individual tests as _filter_discovery_sought,
    _filter_discovery_resolved and _filter_discovery_unresolved.
    
    The new file is sourced by tests 775 and 946 which are fixed
    using the new _filter_discovery_unresolvable.

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