Changes committed to git://oss.sgi.com/kenj/pcp.git dev
qa/common.check | 2 +-
src/libpcp/src/pmns.c | 16 ++++++++++++----
src/libpcp_pmda/src/tree.c | 16 +++++++++++++---
src/pmdas/linux/proc_net_dev.c | 12 +++++++++++-
4 files changed, 37 insertions(+), 9 deletions(-)
commit 485860a0fb6f693bbc3f49fb5fdad251bc1907aa
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Apr 11 08:30:55 2014 +1000
libpcp_pmda - pmdaTreeChildren tweak for dynamic metrics
Tighten the guard to prevent a couple of malloc(0) cases that
were causing qa/957 fallout.
commit 9b752b675678412cd1c77ba3738ef4df35c67b1c
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Apr 11 08:27:27 2014 +1000
libpcp - pmns traversal fixups
Fixed a couple of zero-sized malloc() instances in the PM_CONTEXT_LOCAL
case expanding the PMNS below a dynamic metric. Tightening the guards
prevents the pointless malloc()s.
This was causing some qa/957 fallout.
commit 0ef492a9287f046542e9f61345af32313bf10f98
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Apr 11 06:17:58 2014 +1000
linux PMDA proc_net_dev.c - small initialization change for
valgrind WAR
Comment from the code ...
* Note:
* Initialization of ecmd is not really needed. If the ioctl()s
* work, ecmd is filled in ... but valgrind (at least up to
* version 3.9.0) does not know about the SIOCETHTOOL ioctl()
* and thinks the use of ecmd after this call propagates
* uninitialized data in to ioc.speed and ioc.duplex, causing
* failures for qa/957
commit ecfa750e321149dac484cea68fb9a34029737f5a
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Thu Apr 10 11:47:40 2014 +1000
qa/common.check - "source" is not a sh keyword
I know you really meant "." not "source".
|